Most of the time we'll want to check the connection count of the
criterion we're currently tracking. So instead of duplicating the
src* tests, let's add trk_conn_cnt to report the total number of
connections from the stick table entry currently being tracked.
A nice part of the code was factored, and we should do the same
for the other criteria.
The new "bytes_in_cnt" and "bytes_out_cnt" session counters have been
added. They're automatically updated when session counters are updated.
They can be matched with the "src_kbytes_in" and "src_kbytes_out" ACLs
which apply to the volume per source address. This can be used to deny
access to service abusers.
The new "conn_cur" session counter has been added. It is automatically
updated upon "track XXX" directives, and the entry is touched at the
moment we increment the value so that we don't consider further counter
updates as real updates, otherwise we would end up updating upon completion,
which may not be desired. Probably that some other event counters (eg: HTTP
requests) will have to be updated upon each event though.
This counter can be matched against current session's source address using
the "src_conn_cur" ACL.
It was not normal to have counter fetches in proto_tcp.c. The only
reason was that the key based on the source address was fetched there,
but now we have split the key extraction and data processing, we must
move that to a more appropriate place. Session seems OK since the
counters are all manipulated from here.
Also, since we're precisely counting number of connections with these
ACLs, we rename them src_conn_cnt and src_updt_conn_cnt. This is not
a problem right now since no version was emitted with these keywords.
The "_cnt" suffix is already used by ACLs to count various data,
so it makes sense to use the same one in "conn_cnt" instead of
"conn_cum" to count cumulated connections.
This is not a problem because no version was emitted with those
keywords.
Thus we'll try to stick to the following rules :
xxxx_cnt : cumulated event count for criterion xxxx
xxxx_cur : current number of concurrent entries for criterion xxxx
xxxx_rate: event rate for criterion xxxx
This patch adds the ability to set a pointer in the session to an
entry in a stick table which holds various counters related to a
specific pattern.
Right now the syntax matches the target syntax and only the "src"
pattern can be specified, to track counters related to the session's
IPv4 source address. There is a special function to extract it and
convert it to a key. But the goal is to be able to later support as
many patterns as for the stick rules, and get rid of the specific
function.
The "track-counters" directive may only be set in a "tcp-request"
statement right now. Only the first one applies. Probably that later
we'll support multi-criteria tracking for a single session and that
we'll have to name tracking pointers.
No counter is updated right now, only the refcount is. Some subsequent
patches will have to bring that feature.
This ACL's count can change along the session's life because it depends
on other sessions' activity. Switch it to volatile since any session
could appear while evaluating the ACLs.
The buffer_feed* functions that are used to send data to buffers did only
support sending contiguous chunks while they're relying on memcpy(). This
patch improves on this by making them able to write in two chunks if needed.
Thus, the buffer_almost_full() function has been improved to really consider
the remaining space and not just what can be written at once.
Now we stop relying on BF_READ_DONTWAIT, which is unrelated to the
wakeups, and only consider activity to decide whether to wake the task
up instead of considering the other side's activity. It is worth noting
that the local stream interface's flags were not updated consecutively
to a call to chk_snd(), which could possibly result in hung tasks from
time to time. This fix will avoid possible loops and uncaught events.
Sometimes it's necessary to be able to perform some "layer 6" analysis
in the backend. TCP request rules were not available till now, although
documented in the diagram. Enable them in backend now.
When resetting a session's request analysers, we must take them from the
listener, not from the frontend. At the moment there is no difference
but this might change.
Since the BF_READ_ATTACHED bug was fixed, a new issue surfaced. When
a connection closes on the return path in tunnel mode while the request
input is already closed, the request analyser which is waiting for a
state change never gets woken up so it never closes the request output.
This causes stuck sessions to remain indefinitely.
One way to reliably reproduce the issue is the following (note that the
client expects a keep-alive but not the server) :
server: printf "HTTP/1.0 303\r\n\r\n" | nc -lp8080
client: printf "GET / HTTP/1.1\r\n\r\n" | nc 127.1 2500
The reason for the issue is that we don't wake the analysers up on
stream interface state changes. So the least intrusive and most reliable
thing to do is to consider stream interface state changes to call the
analysers.
We just need to remember what state each series of analysers have seen
and check for the differences. In practice, that works.
A later improvement later could consist in being able to let analysers
state what they're interested to monitor :
- left SI's state
- right SI's state
- request buffer flags
- response buffer flags
That could help having only one set of analysers and call them once
status changes.
After a read, there was a condition to mandatorily wake the task
up if the BF_READ_DONTWAIT flag was set. This was wrong because
the wakeup condition in this case can be deduced from the other
ones. Another condition was put on the other side not being in
SI_ST_EST state. It is not appropriate to do this because it
causes a useless wakeup at the beginning of every first request
in case of speculative polling, due to the fact that we don't
read anything and that the other side is still in SI_ST_INI.
Also, the wakeup was performed whenever to_forward was null,
which causes an unexpected wakeup upon the first read for the
same reason. However, those two conditions are valid if and
only if at least one read was performed.
Also, the BF_SHUTR flag was tested as part of the wakeup condition,
while this one can only be set if BF_READ_NULL is set too. So let's
simplify this ambiguous test by removing the BF_SHUTR part from the
condition to only process events.
Last, the BF_READ_DONTWAIT flag was unconditionally cleared,
while sometimes there would have been no I/O. Now we only clear
it once the I/O operation has been performed, which maintains
its validity until the I/O occurs.
Finally, those fixes saved approximately 16% of the per-session
wakeups and 20% of the epoll_ctl() calls, which translates into
slightly less under high load due to the request often being ready
when the read() occurs. A performance increase between 2 and 5% is
expected depending on the workload.
It does not seem necessary to backport this change to 1.4, eventhough
it fixes some performance issues. It may later be backported if
required to fix something else because the risk of regression seems
very low due to the fact that we're more in line with the documented
semantics.
Some freq counters will have to work on periods different from 1 second.
The original freq counters rely on the period to be exactly one second.
The new ones (freq_ctr_period) let the user define the period in ticks,
and all computations are operated over that period. When reading a value,
it indicates the amount of events over that period too.
This will be used when an I/O handler running in a stream interface
needs to establish a connection somewhere. We want the session
processor to evaluate both I/O handlers, depending on which side has
one. Doing so also requires that stream_int_update_embedded() wakes
the session up only when the other side is established or has closed,
for instance in order to handle connection errors without looping
indefinitely during the connection setup time.
The session processor still relies on BF_READ_ATTACHED being set,
though we must do whatever is required to remove this dependency.
When a connection is closed on a stream interface, some iohandlers
will need to be informed in order to release some resources. This
normally happens upon a shutr+shutw. It is the equivalent of the
fd_delete() call which is done for real sockets, except that this
time we release internal resources.
It can also be used with real sockets because it does not cost
anything else and might one day be useful.
Till now when a server was configured with address 0.0.0.0, the
connection was forwarded to this address which generally is intercepted
by the system as a local address, so this was completely useless.
One sometimes useful feature for outgoing transparent proxies is to
be able to forward the connection to the same address the client
requested. This patch fixes the meaning of 0.0.0.0 precisely to
ensure that the connection will be forwarded to the initial client's
destination address.
(cherry picked from commit 61ba936e6858dfcf9964d25870726621d8188fb9)
[ note: the bug was finally not present in 1.5-dev but at least we
have to reset store_count to be compatible with 1.4 ]
Commit d6e9e3b5e320b957e6c491bd92d91afad30ba638 caused recently created
entries to be removed as soon as they were created, breaking stickiness.
It is not clear whether a use-after-free was possible or not in this case.
This bug was reported by Ben Congleton and narrowed down by Herv Commowick,
both of whom also tested the fix. Thanks to them !
The quote_arg() function can be used to quote an argument or indicate
"end of line" if it's null or empty. It should be useful to more precisely
report location of problems in the configuration.
When an entry already exists, we just need to update its expiration
timer. Let's have a dedicated function for that instead of spreading
open code everywhere.
This change also ensures that an update of an existing sticky session
really leads to an update of its expiration timer, which was apparently
not the case till now. This point needs to be checked in 1.4.
This change makes use of the stick-tables to keep track of any source
address activity. Two ACLs make it possible to check the count of an
entry or update it and act accordingly. The typical usage will be to
reject a TCP request upon match of an excess value.
Till now sticky sessions only held server IDs. Now there are other
data types so it is not acceptable anymore to overwrite the server ID
when writing something. The server ID must then only be written from
the caller when appropriate. Doing this has also led to separate
lookup and storage.
This one can be parsed on the "stick-table" after with the "store"
keyword. It will hold the number of connections matching the entry,
for use with ACLs or anything else.
The stick_tables will now be able to store extra data for a same key.
A limited set of extra data types will be defined and for each of them
an offset in the sticky session will be assigned at startup time. All
of this information will be stored in the stick table.
The extra data types will have to be specified after the new "store"
keyword of the "stick-table" directive, which will reserve some space
for them.
pattern.c depended on stick_table while in fact it should be the opposite.
So we move from pattern.c everything related to stick_tables and invert the
dependency. That way the code becomes more logical and intuitive.
The name 'exps' and 'keys' in struct stksess was confusing because it was
the same name as in the table which holds all of them, while they only hold
one node each. Remove the trailing 's' to more clearly identify who's who.
Right now we're only able to store a server ID in a sticky session.
The goal is to be able to store anything whose size is known at startup
time. For this, we store the extra data before the stksess pointer,
using a negative offset. It will then be easy to cumulate multiple
data provided they each have their own offset.
It's very disturbing to see the "denied req" counter increase without
any other session counter moving. In fact, we can't count a rejected
TCP connection as "denied req" as we have not yet instanciated any
session at all. Let's use a new counter for that.
The frontend's connection was accounted for once the session was
instanciated. This was problematic because the early ACLs weren't
able to correctly account for the number of concurrent connections.
Now we count the connection once it is assigned to the frontend.
It also brings the nice advantage of being more symmetrical, because
the stream_sock's accept() does not have to account for that anymore,
only the session's accept() does.
Now we're able to reject connections very early, so we need to use a
different counter for the connections that are received and the ones
that are accepted and converted into sessions, so that the rate limits
can still apply to the accepted ones. The session rate must still be
used to compute the rate limit, so that we can reject undesired traffic
without affecting the rate.
A new function session_accept() is now called from the lower layer to
instanciate a new session. Once the session is instanciated, the upper
layer's frontent_accept() is called. This one can be service-dependant.
That way, we have a 3-phase accept() sequence :
1) protocol-specific, session-less accept(), which is pointed to by
the listener. It defaults to the generic stream_sock_accept().
2) session_accept() which relies on a frontend but not necessarily
for use in a proxy (eg: stats or any future service).
3) frontend_accept() which performs the accept for the service
offerred by the frontend. It defaults to frontend_accept() which
is really what is used by a proxy.
The TCP/HTTP proxies have been moved to this mode so that we can now rely on
frontend_accept() for any type of session initialization relying on a frontend.
The next step will be to convert the stats to use the same system for the stats.
This will be needed for the last factoring step which adds support
for application-level accept(). The tcp/http accept() code has now
been isolated and will have to move to a separate function.
Till now, the frontend relied on the backend's options for INDEPSTR,
while at the time of accept, the frontend and backend are the same.
So we now use the frontend's pointer instead of the backend and we
don't have any dependency on the backend anymore in the frontend's
accept code.
The conn_retries attribute is now assigned when switching from SI_ST_INI
to SI_ST_REQ. This eliminates one of the last dependencies on the backend
in the frontend's accept() function.
The conn_retries still lies in the session and its initialization depends
on the backend when it may not yet be known. Let's first move it to the
stream interface.
The frontend has no reason to initialize the server-side stream_interface.
It's a leftover from old times which now makes no sense due to the fact
that we don't know in the frontend whether the other side will be a socket,
a task or anything else. Removing this part is possible due to previous
patches which perform the initialization at the proper place. We'll still
have to be able to register an I/O handler for situations where everything
is known only to the frontend (eg: unix stats socket), before merging the
various instanciations of this accept() function.
It's not normal to initialize the server-side stream interface from the
accept() function, because it may change later. Thus, we introduce a new
stream_sock_prepare_interface() function which is called just before the
connect() and which sets all of the stream_interface's callbacks to the
default ones used for real sockets. The ->connect function is also set
at the same instant so that we can easily add new server-side protocols
soon.
It was particularly embarrassing that the server timeout was assigned
to buffers during an accept() just to be potentially changed later in
case of a use_backend rule. The frontend side has nothing to do with
server timeouts.
Now we initialize them right after the connect() succeeds. Later this
should change for a unique stream-interface timeout setting only.
Calling sess_establish() upon a successful connect() was essential, but
it was not clearly stated whether it was necessary for an access to an
I/O handler or not. While it would be desired, having it automatically
add the response analyzers is quite a problem, and it breaks HTTP stats.
The solution is thus not to call it for now and to perform the few response
initializations as needed.
For the long term, we need to find a way to specify the analyzers to install
during a stream_int_register_handler() if any.
The connection timeout stored in the buffer has not been used since the
stream interface were introduced. Let's get rid of it as it's one of the
things that complicate factoring of the accept() functions.
We can disable the monitor-net rules on a listener if this flag is not
set in the listener's options. This will be useful when we don't want
to check that fe->addr is set or not for non-TCP frontends.
The new LI_O_TCP_RULES listener option indicates that some TCP rules
must be checked upon accept on this listener. It is now checked by
the frontend and the L4 rules are evaluated only in this case. The
flag is only set when at least one tcp-req rule is present in the
frontend.
The L4 rules check function has now been moved to proto_tcp.c where
it ought to be.
The tcp inspection rules were fast but were only processed after a
schedule had occurred and all resources were allocated. When defending
against DDoS, it's important to be able to apply some protection the
earliest possible instant.
Thus we introduce a new set of rules : tcp-request rules which act
on pure layer4 information (no content). They are evaluated even
before the buffers are allocated for the session, saving as much
time as possible. That way it becomes possible to check an incoming
connection's source IP address against a list of authorized/blocked
networks, and immediately drop the connection.
The rules are checked even before we perform any socket-specific
operation, so that we can optimize the reject case, which will be the
problematic one during a DDoS. The second stream interface and s->txn
are also now initialized after the rules are parsed for the same
reason. All these optimisations have permitted to reach up to 212000
connnections/s with a real rule rejecting based on the source IP
address.
For a long time we had two large accept() functions, one for TCP
sockets instanciating proxies, and another one for UNIX sockets
instanciating the stats interface.
A lot of code was duplicated and both did not work exactly the same way.
Now we have a stream_sock layer accept() called for either TCP or UNIX
sockets, and this function calls the frontend-specific accept() function
which does the rest of the frontend-specific initialisation.
Some code is still duplicated (session & task allocation, stream interface
initialization), and might benefit from having an intermediate session-level
accept() callback to perform such initializations. Still there are some
minor differences that need to be addressed first. For instance, the monitor
nets should only be checked for proxies and not for other connection templates.
Last, we renamed l->private as l->frontend. The "private" pointer in
the listener is only used to store a frontend, so let's rename it to
eliminate this ambiguity. When we later support detached listeners
(eg: FTP), we'll add another field to avoid the confusion.
The 'client.c' file now only contained frontend-specific functions,
so it has naturally be renamed 'frontend.c'. Same for client.h. This
has also been an opportunity to remove some cross references from
files that should not have depended on it.
In the end, this file should contain a protocol-agnostic accept()
code, which would initialize a session, task, etc... based on an
accept() from a lower layer. Right now there are still references
to TCP.
Some ACLs in the client ought to belong to proto_tcp, or protocols.
This file should only contain frontend-specific information and will
be renamed that way in next commit.
Some functions which act on generic buffer contents without being
tcp-specific were historically in proto_tcp.c. This concerns ACLs
and RDP cookies. Those have been moved away to more appropriate
locations. Ideally we should create some new files for each layer6
protocol parser. Let's do that later.
Right now we count the incoming connection only once everything has
been allocated. Since we're planning on considering early ACL rules,
we need to count the connection earlier.
Just like we do on health checks, we should consider that ACLs that make
use of buffer data are layer 6 and not layer 4, because we'll soon have
to distinguish between pure layer 4 ACLs (without any buffer) and these
ones.
If a "stick store-request" rule is present, an entry is preallocated during
the request. However, if there is no response due to an error or to a redir
mode server, we never release it.
By using msg->sol as the beginning of a message, wrong messages were
displayed in debug mode when they were truncated on the last line,
because msg->sol points to the beginning of the last line. Use
data+msg->som instead.
This would only be wrong when the server has not completely responded yet.
Fix two other occurrences of wrong rsp<->sl associations which were harmless
but wrong anyway.
The rate-limit feature relied on a timer to define how long a frontend
must remain idle. It was not considering the pending connections, so it
was almost always ready to be used again and only the accept's limit was
preventing new connections from coming in. By accounting for the pending
connection, we can compute a correct delay and effectively make the
frontend go idle for that (short) time.
Latest BF_READ_ATTACHED fix has unveiled a nice issue with the way
HTTP requests and responses are forwarded. The case where the request
aborts after the response has responded (POST with early response)
forgot to re-enable auto-close on the response. In fact it still
worked thanks to a side effect as long as BF_READ_ATTACHED was there
to force the states to be resynced (and the flags). Since last fix,
the missing auto-close causes CLOSE_WAIT connections when the client
aborts too late during a data transfer.
The right fix consists in considering the situation where the client
experiences an error and to explicitly abort the transfer. There is
no need to wake the response analysers up for that since they'd have
no added value and the analysers flags are cleared. However for a
future usage, that might help (eg: stickiness, ...).
This fix should be backported to 1.4 if the previous one is backported
too. After all the non-reg tests, the risks to see a problem arise
without both patches seems low, and both patches touch sensible areas
of the code. So there's no hurry.
The BF_READ_ATTACHED flag was created to wake analysers once after
a connection was established. It turns out that this flag is never
cleared once set, so even if there is no event, some analysers are
still evaluated for no reason.
The bug was introduced with commit ea38854d34.
It may cause slightly increased CPU usages during data transfers, maybe
even quite noticeable once when transferring transfer-encoded data,
due to the fact that the request analysers are being checked for every
chunk.
This fix must be backported in 1.4 after all non-reg tests have been
completed.
The response analyser was not emptied upon creation of a new session. In
fact it was always zero just because last session leaved it in a zero state,
but in case of shared pools this cannot be guaranteed. The net effect is
that it was possible to have some HTTP (or any other) analysers on the
response path of a stats unix socket, which would reject the response.
This fix must be backported to 1.4.
Commit 4605e3b641cebbdbb2ee5726e5bbc3c03a2d7b5e was not enough, because
connections passing from a TCP frontend to an HTTP backend without any
ACL and via a "default_backend" statement were still working on non-initialized
data. An initialization was missing in the session_set_backend() function, next
to the initialization of hdr_idx.
It was once reported at least by Dirk Taggesell that the consistent
hash had a very poor distribution, making use of only two servers.
Jeff Persch analysed the code and found the root cause. Consistent
hash makes use of the server IDs, which are completed after the chash
array initialization. This implies that each server which does not
have an explicit "id" parameter will be merged at the same place in
the chash tree and that in the end, only the first or last servers
may be used.
The now obvious fix (thanks to Jeff) is to assign the missing IDs
earlier. However, it should be clearly understood that changing a
hash algorithm on live systems will rebalance the whole system.
Anyway, the only affected users will be the ones for which the
system is quite unbalanced already. The ones who fix their IDs are
not affected at all.
Kudos to Jeff for spotting that bug which got merged 3 days after
the consistent hashing !
When running in pure TCP mode with a traffic inspection rule to detect
HTTP protocol, we have to initialize the HTTP transaction too. The
effect of not doing this was that some incoming connections could have
been matched as carrying HTTP protocol eventhough this was not the case.
Both dispatch and http_proxy modes were broken since 1.4-dev5 when
the adjustment of server health based on response codes was introduced.
In fact, in these modes, s->srv == NULL. The result is a plain segfault.
It should have been noted critical, but the fact that it remained 6
months without being noticed indicates that almost nobody uses these
modes anymore. Also, the crash is immediate upon first request.
Further versions should not be affected anymore since it's planned to
have a dummy server instead of these annoying NULL pointers.
This ACL was missing in complex setups where the status of a remote site
has to be considered in switching decisions. Until there, using a server's
status in an ACL required to have a dedicated backend, which is a bit heavy
when multiple servers have to be monitored.
It is now possible to stick on an IP address found in a HTTP header. Right
now only the last occurrence of the header can be used, which is generally
enough for most uses. Also, the header extraction rule only knows how to
convert the header to IP. Later it will be usable as a plain string with
an implicit conversion, and the syntax will not change.
Most often, pattern files used by ACLs will be produced by tools
which emit some comments (eg: geolocation lists). It's very annoying
to have to clean the files before using them, and it does not make
much sense to be able to support patterns we already can't input in
the config file. So this patch makes the pattern file loader skip
lines beginning with a sharp and the empty ones, and strips leading
spaces and tabs.
Networks patterns loaded from files for longest match ACL testing
will now be arranged into a prefix tree. This is possible thanks to
the new prefix features in ebtree v6.0. Longest match testing is
slightly slower than exact data maching. However, the measured impact
of running at 42000 requests per second and testing whether the IP
address found in a header belongs to a list of 52000 networks or
not is 3% CPU (increase from 66% to 69%). This is low enough to
permit true geolocation based on huge tables.
Now if some ACL patterns are loaded from a file and the operation is
an exact string match, the data will be arranged in a tree, yielding
a significant performance boost on large data sets. Note that this
only works when case is sensitive.
A new dedicated function, acl_lookup_str(), has been created for this
matching. It is called for every possible input data to test and it
looks the tree up for the data. Since the keywords are loosely typed,
we would have had to add a new columns to all keywords to adjust the
function depending on the type. Instead, we just compare on the match
function. We call acl_lookup_str() when we could use acl_match_str().
The tree lookup is performed first, then the remaining patterns are
attempted if the tree returned nothing.
A quick test shows that when matching a header against a list of 52000
network names, haproxy uses 68% of one core on a core2-duo 3.2 GHz at
42000 requests per second, versus 66% without any rule, which means
only a 2% CPU increase for 52000 rules. Doing the same test without
the tree leads to 100% CPU at 6900 requests/s. Also it was possible
to run the same test at full speed with about 50 sets of 52000 rules
without any measurable performance drop.
The code is now ready to support loading pattern from filesinto trees. For
that, it will be required that the ACL keyword has a flag ACL_MAY_LOOKUP
and that the expr is case sensitive. When that is true, the pattern will
have a flag ACL_PAT_F_TREE_OK to indicate that it is possible to feed the
tree instead of a usual pattern if the parsing function is able to do this.
The tree's root is pre-initialized in the pattern's value so that the
function can easily find it. At that point, if the parsing function decides
to use the tree, it just sets ACL_PAT_F_TREE in the return flags so that
the caller knows the tree has been used and the pattern can be recycled.
That way it will be possible to load some patterns into the tree when it
is compatible, and other ones as linear linked lists. A good example of
this might be IPv4 network entries : right now we support holes in masks,
but this very rare feature is not compatible with binary lookup in trees.
So the parser will be able to decide itself whether the pattern can go to
the tree or not.
The "acl XXX -f <file>" syntax was supported but nothing was read from
the file. This is now possible. All lines are merged verbatim, even if
they contain spaces (useful for user-agents). There are shortcomings
though. The worst one is that error reporting is too approximative.
in cttproxy.c check_cttproxy_version socket is not closed before function
returned. Although it is called only once, I think it is better to close
the socket.
When trying to display an invalid request or response we received,
we must at least check that we have identified something looking
like a start of message, otherwise we can dereference a NULL pointer.
This is used to disable persistence depending on some conditions (for
example using an ACL matching static files or a specific User-Agent).
You can see it as a complement to "force-persist".
In the configuration file, the force-persist/ignore-persist declaration
order define the rules priority.
Used with the "appsesion" keyword, it can also help reducing memory usage,
as the session won't be hashed the persistence is ignored.
I met a strange behaviour with appsession.
I firstly thought this was a regression due to one of my previous patch
but after testing with a 1.3.15.12 version, I also could reproduce it.
To illustrate, the configuration contains :
appsession PHPSESSID len 32 timeout 1h
Then I call a short PHP script containing :
setcookie("P", "should not match")
When calling this script thru haproxy, the cookie "P" matches the appsession rule :
Dumping hashtable 0x11f05c8
table[1572]: should+not+match
Shouldn't it be ignored ?
If you confirm, I'll send a patch for 1.3 and 1.4 branches to check that the
cookie length is equal to the appsession name length.
This is due to the comparison length, where the cookie length is took into
account instead of the appsession name length. Using the appsession name
length would allow ASPSESSIONIDXXX (+ check that memcmp won't go after the
buffer size).
Also, while testing, I noticed that HEAD requests where not available for
URIs containing the appsession parameter. 1.4.3 patch fixes an horrible
segfault I missed in a previous patch when appsession is not in the
configuration and HAProxy is compiled with DEBUG_HASH.
Some servers do not completely conform with RFC2616 requirements for
keep-alive when they receive a request with "Connection: close". More
specifically, they don't bother using chunked encoding, so the client
never knows whether the response is complete or not. One immediately
visible effect is that haproxy cannot maintain client connections alive.
The second issue is that truncated responses may be cached on clients
in case of network error or timeout.
scar Fras Barranco reported this issue on Tomcat 6.0.20, and
Patrik Nilsson with Jetty 6.1.21.
Cyril Bont proposed this smart idea of pretending we run keep-alive
with the server and closing it at the last moment as is already done
with option forceclose. The advantage is that we only change one
emitted header but not the overall behaviour.
Since some servers such as nginx are able to close the connection
very quickly and save network packets when they're aware of the
close negociation in advance, we don't enable this behaviour by
default.
"option http-pretend-keepalive" will have to be used for that, in
conjunction with "option http-server-close".
Using get_ip_from_hdr2() we can look for occurrence #X or #-X and
extract the IP it contains. This is typically designed for use with
the X-Forwarded-For header.
Using "usesrc hdr_ip(name,occ)", it becomes possible to use the IP address
found in <name>, and possibly specify occurrence number <occ>, as the
source to connect to a server. This is possible both in a server and in
a backend's source statement. This is typically used to use the source
IP previously set by a upstream proxy.
The transparent proxy address selection was set in the TCP connect function
which is not the most appropriate place since this function has limited
access to the amount of parameters which could produce a source address.
Instead, now we determine the source address in backend.c:connect_server(),
right after calling assign_server_address() and we assign this address in
the session and pass it to the TCP connect function. This cannot be performed
in assign_server_address() itself because in some cases (transparent mode,
dispatch mode or http_proxy mode), we assign the address somewhere else.
This change will open the ability to bind to addresses extracted from many
other criteria (eg: from a header).
We'll need another flag in the 'options' member close to PR_O_TPXY_*,
and all are used, so let's move this easy one to options2 (which are
already used for SQL checks).
It's very common to see people getting trapped by HTTP-only options
which don't work in TCP proxies. To help them definitely get rid of
those configs, let's emit warnings for all options and statements
which are not supported in their mode. That includes all HTTP-only
options, the cookies and the stats.
In order to ensure internal config correctness, the options are also
disabled.
It was disturbing to see a backend name associated with a bad request
when this "backend" was in fact the frontend. Instead, we now display
"backend <NONE>" if the "backend" has no backend capability :
> show errors
[25/Mar/2010:06:44:25.394] frontend fe (#1): invalid request
src 127.0.0.1, session #0, backend <NONE> (#-1), server <NONE> (#-1)
request length 45 bytes, error at position 0:
Isidore Li reported an occasional segfault when using URL hashing, and
kindly provided backtraces and core files to help debugging.
The problem was triggered by reset connections before the URL was sent,
and was due to the same bug which was fixed by commit e45997661b
(connections were attempted in case of connection abort). While that
bug was already fixed, it appeared that the same segfault could be
triggered when URL hashing is configured in an HTTP backend when the
frontend runs in TCP mode and no URL was seen. It is totally abnormal
to try to hash a null URL, as well as to process any kind of L7 hashing
when a full request was not seen.
This additional fix now ensures that layer7 hashing is not performed on
incomplete requests.
The following patch fixed an issue but brought another one :
296897 [MEDIUM] connect to servers even when the input has already been closed
The new issue is that when a connection is inspected and aborted using
TCP inspect rules, now it is sent to the server before being closed. So
that test is not satisfying. A probably better way is not to prevent a
connection from establishing if only BF_SHUTW_NOW is set but BF_SHUTW
is not. That way, the BF_SHUTW flag is not set if the request has any
data pending, which still fixes the stats issue, but does not let any
empty connection pass through.
Also, as a safety measure, we extend buffer_abort() to automatically
disable the BF_AUTO_CONNECT flag. While it appears to always be OK,
it is by pure luck, so better safe than sorry.
This warning was first reported by Ross West on FreeBSD, then by
Holger Just on OpenSolaris. It also happens on 64bit Linux. However,
fixing the format to use long int complains on 32bit Linux where
ptrdiff_t is apparently different. Better cast the pointer difference
to an int then.
To save a little memory, the check_data buffer is only allocated
for the servers that are checked.
[WT: this patch saves 80 MB of RAM on the test config with 5000 servers]
Cyril Bont reported a regression introduced with very last changes
on the checks code, which causes failed checks on if the server does
not close the connection in time. This happens on HTTP/1.1 checks or
on SMTP checks for instance.
This fix consists in restoring the old behaviour of parsing as soon
as something is available in the response buffer, and waiting for
more data if some are missing. This also helps releasing connections
earlier (eg: a GET check will not have to download the whole object).
Apparently some systems define MSG_NOSIGNAL but do not necessarily
check it (or maybe binaries are built somewhere and used on older
versions). There were reports of very recent FreeBSD setups causing
SIGPIPEs, while older ones catch the signal. Recent FreeBSD manpages
indeed define MSG_NOSIGNAL.
So let's now unconditionnaly catch the signal. It's useless not to do
it for the rare cases where it's not needed (linux 2.4 and below).
Bernhard Krieger reported truncated HTTP responses in presence of some
specific chunk-encoded data, and kindly offered complete traces of the
issue which made it easy to reproduce it.
Those traces showed that the chunks were of exactly 8192 bytes, chunk
size and CRLF included, which was exactly half the size of the buffer.
In this situation, the function http_chunk_skip_crlf() could erroneously
try to parse a CRLF after the chunk believing there were more data
pending, because the number of bytes present in the buffer was considered
instead of the number of remaining bytes to be parsed.
This happens when a server immediately closes the connection after
the response without lingering or when we close before the end of
the data. We get an RST which translates into a late error. We must
not declare an error without checking that the contents are OK.
Since the recv() call returns every time it succeeds, we always need
to calls with one intermediate poll before detecting the end of response :
20:20:03.958207 recv(7, "HTTP/1.1 200\r\nConnection: close\r\n"..., 8030, 0) = 145
20:20:03.958365 epoll_wait(3, {{EPOLLIN, {u32=7, u64=7}}}, 8, 1000) = 1
20:20:03.958543 gettimeofday({1268767203, 958626}, NULL) = 0
20:20:03.958694 recv(7, ""..., 7885, 0) = 0
20:20:03.958833 shutdown(7, 2 /* send and receive */) = 0
Let's read as long as we can, that way we can detect end of connections
in the same call, which is much more efficient especially for LBs with
hundreds of servers :
20:29:58.797019 recv(7, "HTTP/1.1 200\r\nConnection: close\r\n"..., 8030, 0) = 145
20:29:58.797182 recv(7, ""..., 7885, 0) = 0
20:29:58.797356 shutdown(7, 2 /* send and receive */) = 0
We are seeing both real servers repeatedly going on- and off-line with
a period of tens of seconds. Packet tracing, stracing, and adding
debug code to HAProxy itself has revealed that the real servers are
always responding correctly, but HAProxy is sometimes receiving only
part of the response.
It appears that the real servers are sending the test page as three
separate packets. HAProxy receives the contents of one, two, or three
packets, apparently randomly. Naturally, the health check only
succeeds when all three packets' data are seen by HAProxy. If HAProxy
and the real servers are modified to use a plain HTML page for the
health check, the response is in the form of a single packet and the
checks do not fail.
(...)
I've added buffer and length variables to struct server, and allocated
space with the rest of the server initialisation.
(...)
It seems to be working fine in my tests, and handles check responses
that are bigger than the buffer.
Those two codes can be triggered on demand by client requests.
We must not fail a server on them.
Ideally we should ignore a certain amount of status codes which do
not indicate life nor death.
The new anonymous ACL feature was buggy. If several ones are
declared, the first rule is always matched because all of them
share the same internal name (".noname"). Now we simply declare
them with an empty name and ensure that we disable any merging
when the name is empty.
Hi Willy,
Please find a small patch to prevent haproxy segfaulting when logging captured headers in CLF format.
Example config to reproduce the bug :
listen test :10080
log 127.0.0.1 local7 debug err
mode http
option httplog clf
capture request header NonExistantHeader len 16
--
Cyril Bont
The BF_AUTO_CLOSE flag prevented a connection from establishing on
a server if the other side's input channel was already closed. This
is wrong because there may be pending data to be sent.
This was causing an issue with stats, as noticed and reported by
Cyril Bont. Since the stats are now handled as a server, sometimes
concurrent accesses were causing one of the connections to send the
shutdown(write) before the connection to the stats function was
established, which aborted it early.
This fix causes the BF_AUTO_CLOSE flag to be checked only when the
connection on the outgoing stream interface has reached an established
state. That way we're still able to connect, send the request then
close.
Marcello Gorlani reported that at least on FreeBSD, a long hostname
was reported with garbage on the stats page. POSIX does not make it
mandatory for gethostname() to NULL-terminate the string in case of
truncation, and at least FreeBSD appears not to do it. So let's
force null-termination to keep safe.
Since the last documentation cleanups, I've found more typos that I kept
in a corner instead of sending you a mail just for one character :)
--
Cyril Bont
Jozef Hovan reported a bug sometimes causing a down server to be
used in url_param hashing mode.
This happens if the following conditions are met :
- the backend contains more than one server with at least two
of different weights
- all servers but one are down
- the server which is not down has a weight which does not divide
all the other ones
Example: 3 servers with 20,20,10, the first one remains up.
The problem is caused by an optimisation in recalc_server_map()
which only fills the first map slot when only one server is up,
because all LB algorithms are optimized to use entry zero when
only one server is up... All but url_param. When doing the modulus,
we can return a position which is greater than zero and use an
entry which still refers to a server which has since been stopped.
One solution could be to optimize the url_param algo to proceed
as the other ones, but the fact that was wrong implies that we
can repeat the same bug later. So let's first correctly initialize
the map in order to avoid that trap.
When trying to spot some complex bugs, it's often needed to access
information on stuck sessions, which is quite difficult. This new
command helps one get detailed information about a session, with
flags, timers, states, etc... The buffer data are not dumped yet.
In case of pipelined requests, if the client aborts before reading response
N-1, haproxy waits forever for the data to leave the buffer before parsing
the next response.
This duplicate test should have been removed with the loop rework but was forgotten.
It was harmless, but disassembly shows that it prevents gcc from correctly optimizing
the loop.
We have been using haproxy to balance a not very well written application
(http://www.blackboard.com/). Using the "insert postonly indirect" cookie
method, I was attempting to remove the cookie when users would logout,
allowing the machine to re-balance for the next user (this application is
used in school computer labs, so a computer might stay on the whole day
but be used on and off).
I was having a lot of trouble because when the cookie was set, it was with
"Path=/", but when being cleared there was no "Path" in the set cookie
header, and because the logout page was in a different place of the
website (which I couldn't change), the cookie would not be cleared. I
don't know if this would be a problem for anyone other than me (as our
HTTP application is so un-adjustable), but just in case, I have included
the patch I used. Maybe it will help someone else.
[ WT: this was a correct fix, and I also added the same missing path to
the set-cookie option ]
It's sometimes convenient to know if a proxy has processed any connection
at all when stopping it. Since a soft restart causes the "Proxy stopped"
message to be logged for each proxy, let's add the number of connections
so that it's possible afterwards to check whether a proxy had received
any connection.
Often we need to understand why some transfers were aborted or what
constitutes server response errors. With those two counters, it is
now possible to detect an unexpected transfer abort during a data
phase (eg: too short HTTP response), and to know what part of the
server response errors may in fact be assigned to aborted transfers.
Holger Just and Ross West reported build issues on FreeBSD and
Solaris that were initially caused by the definition of
_XOPEN_SOURCE at the top of auth.c, which was required on Linux
to avoid a build warning.
Krzysztof Oledzki found that using _GNU_SOURCE instead also worked
on Linux and did not cause any issue on several versions of FreeBSD.
Solaris still reported a warning this time, which was fixed by
including <crypt.h>, which itself is not present on FreeBSD nor on
all Linux toolchains.
So by adding a new build option (NEED_CRYPT_H), we can get Solaris
to get crypt() working and stop complaining at the same time, without
impacting other platforms.
This fix was tested at least on several linux toolchains (at least
uclibc, glibc 2.2.5, 2.3.6 and 2.7), on FreeBSD 4 to 8, Solaris 8
(which needs crypt.h), and AIX 5.3 (without crypt.h).
Every time it builds without a warning.
A copy-paste typo and a missing check were causing the logs to
report "PR" instead of "SD" when a server closes before sending
full data. Also, the log would erroneously report 502 while in
fact the correct response will already have been transmitted.
Some people have reported seeing "SL" flags in their logs quite often while
this should never happen. The reason was that then a server error is detected,
we close the connection to that server and when we decide what state we were
in, we see the connection is closed, and deduce it was the last data transfer,
which is wrong. We should report DATA if the previous state was an established
state, which this patch does.
Now logs correctly report "SD" and not "SL" when a server resets a connection
before the end of the transfer.
isalnum, isdigit and friends are really annoying because they take
an int in which we should pass an unsigned char, while strings
everywhere use chars. Solaris uses macros relying on an array for
those functions, which easily triggers some warnings showing where
we have mistakenly passed a char instead of an unsigned char or an
int. Those warnings may indicate real bugs on some platforms
depending on the implementation.
There are many information available in the stats page that can only
be seen when the mouse hovers over them. But it's hard to know where
those information are. Now with a discrete dotted underline it's easier
to spot those areas.
The current and max request rates are now reported when the mouse flies
over the session rate cur/max. The total requests is displayed with the
status codes over the total sessions cell.
The bounce realign function was algorithmically good but as expected
it was not cache-friendly. Using it with large requests caused so many
cache thrashing that the function itself could drain 70% of the total
CPU time for only 0.5% of the calls !
Revert back to a standard memcpy() using a specially allocated swap
buffer. We're now back to 2M req/s on pipelined requests.
It is wrong to merge FE and BE stats for a proxy because when we consult a
BE's stats, it reflects the FE's stats eventhough the BE has received no
traffic. The most common example happens with listen instances, where the
backend gets credited for all the trafic even when a use_backend rule makes
use of another backend.
The trash buffer may now be smaller than a buffer because we can tune
it at run time. This causes a risk when we're trying to use it as a
temporary buffer to realign unaligned requests, because we may have to
put up to a full buffer into it.
Instead of doing a double copy, we're now relying on an open-coded
bouncing copy algorithm. The principle is that we move one byte at
a time to its final place, and if that place also holds a byte, then
we move it too, and so on. We finish when we've moved all the buffer.
It limits the number of memory accesses, but since it proceeds one
byte at a time and with random walk, it's not cache friendly and
should be slower than a double copy. However, it's only used in
extreme situations and the difference will not be noticeable.
It has been extensively tested and works reliably.
When a host name could not be resolved, an alert was emitted but the
service used to start with 0.0.0.0 for the IP address, because the
address parsing functions could not report an error. This is now
changed. This fix must be backported to 1.3 as it was first discovered
there.
[WT: it was not a bug, I did it on purpose to leave no hole between IDs,
though it's not very practical when admins want to force some entries
after they have been used, because they'd rather leave a hole than
renumber everything ]
Forcing some of IDs should not shift others.
Regression introduced in 53fb4ae261
---cut here---
global
stats socket /home/ole/haproxy.stat user ole group ole mode 660
frontend F1
bind 127.0.0.1:9999
mode http
backend B1
mode http
backend B2
mode http
id 9999
backend B3
mode http
backend B4
mode http
---cut here---
Before 53fb4ae261:
$ echo "show stat" | socat unix-connect:/home/ole/haproxy.stat stdio|cut -d , -f 28
iid
1
2
9999
4
5
After 53fb4ae261:
$ echo "show stat" | socat unix-connect:/home/ole/haproxy.stat stdio|cut -d , -f 28
iid
1
2
9999
3
4
With this patch:
$ echo "show stat" | socat unix-connect:/home/ole/haproxy.stat stdio|cut -d , -f 28
iid
1
2
9999
4
5
Thich patch fixes cfgparser not to leak memory on each
default server statement and adds several missing free
calls in deinit():
- free(l->name)
- free(l->counters)
- free(p->desc);
- free(p->fwdfor_hdr_name);
None of them are critical, hopefully.
Released version 1.4-rc1 with the following main changes :
- [MEDIUM] add a maintenance mode to servers
- [MINOR] http-auth: last fix was wrong
- [CONTRIB] add base64rev-gen.c that was used to generate the base64rev table.
- [MINOR] Base64 decode
- [MINOR] generic auth support with groups and encrypted passwords
- [MINOR] add ACL_TEST_F_NULL_MATCH
- [MINOR] http-request: allow/deny/auth support for frontend/backend/listen
- [MINOR] acl: add http_auth and http_auth_group
- [MAJOR] use the new auth framework for http stats
- [DOC] add info about userlists, http-request and http_auth/http_auth_group acls
- [STATS] make it possible to change a CLI connection timeout
- [BUG] patterns: copy-paste typo in type conversion arguments
- [MINOR] pattern: make the converter more flexible by supporting void* and int args
- [MINOR] standard: str2mask: string to netmask converter
- [MINOR] pattern: add support for argument parsers for converters
- [MINOR] pattern: add the "ipmask()" converting function
- [MINOR] config: off-by-one in "stick-table" after list of converters
- [CLEANUP] acl, patterns: make use of my_strndup() instead of malloc+memcpy
- [BUG] restore accidentely removed line in last patch !
- [MINOR] checks: make the HTTP check code add the CRLF itself
- [MINOR] checks: add the server's status in the checks
- [BUILD] halog: make without arch-specific optimizations
- [BUG] halog: fix segfault in case of empty log in PCT mode (cherry picked from commit fe362fe476)
- [MINOR] http: disable keep-alive when process is going down
- [MINOR] acl: add build_acl_cond() to make it easier to add ACLs in config
- [CLEANUP] config: use build_acl_cond() instead of parse_acl_cond()
- [CLEANUP] config: use warnif_cond_requires_resp() to check for bad ACLs
- [MINOR] prepare req_*/rsp_* to receive a condition
- [CLEANUP] config: specify correct const char types to warnif_* functions
- [MEDIUM] config: factor out the parsing of 20 req*/rsp* keywords
- [MEDIUM] http: make the request filter loop check for optional conditions
- [MEDIUM] http: add support for conditional request filter execution
- [DOC] add some build info about the AIX platform (cherry picked from commit e41914c77e)
- [MEDIUM] http: add support for conditional request header addition
- [MEDIUM] http: add support for conditional response header rewriting
- [DOC] add some missing ACLs about response header matching
- [MEDIUM] http: add support for proxy authentication
- [MINOR] http-auth: make the 'unless' keyword work as expected
- [CLEANUP] config: use build_acl_cond() to simplify http-request ACL parsing
- [MEDIUM] add support for anonymous ACLs
- [MEDIUM] http: switch to tunnel mode after status 101 responses
- [MEDIUM] http: stricter processing of the CONNECT method
- [BUG] config: reset check request to avoid double free when switching to ssl/sql
- [MINOR] config: fix too large ssl-hello-check message.
- [BUG] fix error response in case of server error
The fix below was incomplete :
commit d5fd51c75b
[BUG] http_server_error() must not purge a previous pending response
This can cause parts of responses to be truncated in case of
pipelined requests if the second request generates an error
before the first request is completely flushed.
Pending response data being rejected was still sent, causing inappropriate
error responses in case of error while parsing a response header. We must
purge pending data from the response buffer that were not scheduled to be
sent (l - send_max).
SSL and SQL checks did only perform a free() of the request without replacing
it, so having multiple SSL/SQL check declarations after another check type
causes a double free condition during config parsing. This should be backported
although it's harmless.
Now we establish the tunnel only once the status 200 reponse is
received. That way we can still support an authentication request
in response to a CONNECT, then a client's authentication response.
A 101 response is accompanied with an Upgrade header indicating
a new protocol that is spoken on the connection after the exchange
completes. At least we should switch to tunnel mode after such a
response.
Anonymous ACLs allow the declaration of rules which rely directly on
ACL expressions without passing via the declaration of an ACL. Example :
With named ACLs :
acl site_dead nbsrv(dynamic) lt 2
acl site_dead nbsrv(static) lt 2
monitor fail if site_dead
With anonymous ACLs :
monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
I'm not sure if the fix is correct:
- if (req_acl->cond)
- ret = acl_exec_cond(req_acl->cond, px, s, txn, ACL_DIR_REQ);
+ if (!req_acl->cond)
+ continue;
Doesn't it ignore rules with no condition attached? I think that the
proper solution would be the following.
One check was missing for the 'polarity' of the test. Now 'unless'
works. BTW, 'unless' provides a nice way to perform one-line auth :
acl valid-user http_auth(user-list)
http-request auth unless valid-user
This is a first attempt to add a maintenance mode on servers, using
the stat socket (in admin level).
It can be done with the following command :
- disable server <backend>/<server>
- enable server <backend>/<server>
In this mode, no more checks will be performed on the server and it
will be marked as a special DOWN state (MAINT).
If some servers were tracking it, they'll go DOWN until the server
leaves the maintenance mode. The stats page and the CSV export also
display this special state.
This can be used to disable the server in haproxy before doing some
operations on this server itself. This is a good complement to the
"http-check disable-on-404" keyword and works in TCP mode.
We're already able to know if a request is a proxy request or a
normal one, and we have an option "http-use-proxy-header" which states
that proxy headers must be checked. So let's switch to use the proxy
authentication headers and responses when this option is set and we're
facing a proxy request. That allows haproxy to enforce auth in front
of a proxy.
Support the new syntax (http-request allow/deny/auth) in
http stats.
Now it is possible to use the same syntax is the same like in
the frontend/backend http-request access control:
acl src_nagios src 192.168.66.66
acl stats_auth_ok http_auth(L1)
stats http-request allow if src_nagios
stats http-request allow if stats_auth_ok
stats http-request auth realm LB
The old syntax is still supported, but now it is emulated
via private acls and an aditional userlist.
Add generic authentication & authorization support.
Groups are implemented as bitmaps so the count is limited to
sizeof(int)*8 == 32.
Encrypted passwords are supported with libcrypt and crypt(3), so it is
possible to use any method supported by your system. For example modern
Linux/glibc instalations support MD5/SHA-256/SHA-512 and of course classic,
DES-based encryption.
Implement Base64 decoding with a reverse table.
The function accepts and decodes classic base64 strings, which
can be composed from many streams as long each one is properly
padded, for example: SGVsbG8=IEhBUHJveHk=IQ==
Just as for the req* rules, we can now condition rsp* rules with ACLs.
ACLs match on response, so volatile request information cannot be used.
A warning is emitted if a configuration contains such an anomaly.
All the req* rules except the reqadd rules can now be specified with
an if/unless condition. If a condition is specified and does not match,
the filter is ignored. This is particularly useful with reqidel, reqirep
and reqtarpit.
From now on, if request filters have ACLs defined, these ACLs will be
evaluated to condition the filter. This will be used to conditionally
remove/rewrite headers based on ACLs.
A new function was added to take care of the common code between
all those keywords. This has saved 8 kB of object code and about
500 lines of source code. This has also permitted to spot and fix
minor bugs (allocated args that were never used).
The code could be factored even more but that would make it a bit
more complex which is not interesting at this stage.
Various tests have been performed, and the warnings and errors are
still correctly reported and everything seems to work as expected.
This function automatically builds a rule, considering the if/unless
statements, and automatically updates the proxy's acl_requires, the
condition's file and line.
Krzysztof Oledzki suggested to disable keep-alive when a process
is going down due to a reload, in order to avoid ever-lasting
sessions. This is a simple and very efficient solution as it
ensures that at most one more request will be handled on a
keep-alive connection after the process has received a SIGUSR1
signal.
Now a server can check the contents of the header X-Haproxy-Server-State
to know how haproxy sees it. The same values as those reported in the stats
are provided :
- up/down status + check counts
- throttle
- weight vs backend weight
- active sessions vs backend sessions
- queue length
- haproxy node name
Currently we cannot easily add headers nor anything to HTTP checks
because the requests are pre-formatted with the last CRLF. Make the
check code add the CRLF itself so that we can later add useful info.
This converter can be applied on top of an IPv4-type pattern. It
applies a netmask which is suited for IP address storage and matching.
This can be used to make all hosts within a certain mask to share the
same table entries and as such use the same server.
The mask can be passed in dotted form (eg: 255.255.255.0) or in CIDR
form (eg: 24).
Some converters will need one or several arguments. It's not possible
to write a simple generic parser for that, so let's add the ability
for each converter to support its own argument parser, and call it
to get the arguments when it's specified. If unspecified, the arguments
are passed unmodified as string+len.
The pattern type converters currently support a string arg and a length.
Sometimes we'll prefer to pass them a list or a structure. So let's convert
the string and length into a generic void* and int that each converter may
use as it likes.
Hi Willy,
I've made a quick pass on the "defaults" column in the Proxy keywords matrix (chapter 4.1. in the documentation).
This patch resyncs the code and the documentation. I let you decide if some keywords that still work in the "defaults" section should be forbidden.
- default_backend : in the matrix, "defaults" was not supported but the keyword details say it is.
Tests also shows it works, then I've updated the matrix.
- capture cookie : in the keyword details, we can read `It is not possible to specify a capture in a "defaults" section.'.
Ok, even if the tests worked, I've added an alert in the configuration parser (as it is for capture request/response header).
- description : not supported in "defaults", I added an alert in the parser.
I've also noticed that this keyword doesn't appear in the documentation.
There's one "description" entry, but for the "global" section, which is for a different use (the patch doesn't update the documentation).
- grace : even if this is maybe useless, it works in "defaults". Documentation is updated.
- redirect : alert is added in the parser.
- rsprep : alert added in the parser.
--
Cyril Bont
We must trim any excess data from the response buffer when recycling
a keep-alive connection, because we may have blocked an invalid response
from a server that we don't want to accidentely forward once we disable
the analysers, nor do we want those data to come along with next response.
A typical example of such data would be from a buggy server responding to
a HEAD with some data, or sending more than the advertised content-length.
For deciding to set the BF_EXPECT_MORE, we reused the same code as in
http_wait_for_request(), but here we must ignore buf->lr which is not
yet set and useless. This might only have caused random sub-optimal
behaviours.
Krzysztof Oledzki reported that 1.4-dev7 would regularly crash
on an apparently very common workload. The cores he provided
showed some inter-buffer data corruption, exactly similar to
what was fixed by the following recent commit :
bbfa7938bd [BUG] buffer_replace2 must never change the ->w entry
In fact, it was buffer_insert_line2() which was still modifying the
->w pointer, causing issues with pipelined responses in keep-alive
mode if some headers were to be added.
The bug requires a remote client, a near server, large server buffers
and small client buffers to be reproduced, with response header
insertion. Still, it's surprizing that it did not trigger earlier.
Now after 100k pipelined requests it did not trigger anymore.
Despite what is explicitly stated in HTTP specifications,
browsers still use the undocumented Proxy-Connection header
instead of the Connection header when they connect through
a proxy. As such, proxies generally implement support for
this stupid header name, breaking the standards and making
it harder to support keep-alive between clients and proxies.
Thus, we add a new "option http-use-proxy-header" to tell
haproxy that if it sees requests which look like proxy
requests, it should use the Proxy-Connection header instead
of the Connection header.
This function is used to move data which is located between ->w and ->r,
so it must not touch ->w, otherwise it will displace pending data which
is before the one we're actually overwriting. The issue arises with
some pipelined responses which cause some part of the previous one to
be chopped off when removing the connection: close header, thus
corrupting last response and shifting next one. Those are detected
in the logs because the next response will be a 502 with flags PH.
When using "option persist" or "force-persist", we want to know from the
logs if the cookie referenced a valid server or a down server. Till here
the flag reported a valid server even if the server was down, which is
misleading. Now we correctly report that the requested server was down.
We can typically see "--DI" when using "option persist" with redispatch,
ad "SCDN" when using force-persist on a down server.
This is used to force access to down servers for some requests. This
is useful when validating that a change on a server correctly works
before enabling the server again.
We use to delay the response if there is a new request in the buffer.
However, if the pending request is incomplete, we should not delay the
pending responses.
This can cause parts of responses to be truncated in case of
pipelined requests if the second request generates an error
before the first request is completely flushed.
Sometimes we need to be able to change the default kernel socket
buffer size (recv and send). Four new global settings have been
added for this :
- tune.rcvbuf.client
- tune.rcvbuf.server
- tune.sndbuf.client
- tune.sndbuf.server
Those can be used to reduce kernel memory footprint with large numbers
of concurrent connections, and to reduce risks of write timeouts with
very slow clients due to excessive kernel buffering.
This one is the next step of previous patch. It correctly computes
the response mode and the Connection flag transformations depending
on the request mode and version, and the response version and headers.
We're now also able to add "Connection: keep-alive", and to convert
server's close during a keep-alive connection to a server-close
connection.
We need to improve Connection header handling in the request for it
to support the upcoming keep-alive mode. Now we have two flags which
keep in the session the information about the presence of a
Connection: close and a Connection: keep-alive headers in the initial
request, as well as two others which keep the current state of those
headers so that we don't have to parse them again. Knowing the initial
value is essential to know when the client asked for keep-alive while
we're forcing a close (eg in server-close mode). Also the Connection
request parser is now able to automatically remove single header values
at the same time they are parsed. This provides greater flexibility and
reliability.
All combinations of listen/front/back in all modes and with both
1.0 and 1.1 have been tested.
Some header values might be delimited with spaces, so it's not enough to
compare "close" or "keep-alive" with strncasecmp(). Use word_match() for
that.
Calling this function after http_find_header2() automatically deletes
the current value of the header, and removes the header itself if the
value is the only one. The context is automatically adjusted for a
next call to http_find_header2() to return the next header. No other
change nor test should be made on the transient context though.
The close mode of a transaction would be switched to tunnel mode
at the end of the processing, letting a lot of pending data pass
in the other direction if any. Let's fix that by checking for the
close mode during state resync too.
We must set the error flags when detecting that a client has reset
a connection or timed out while waiting for a new request on a keep-alive
connection, otherwise process_session() sets it itself and counts one
request error.
That explains why some sites were showing an increase in request errors
with the keep-alive.
We get a lot of those, especially with web crawlers :
recv(2, 0x810b610, 7000, 0) = -1 ECONNRESET (Connection reset by peer)
shutdown(2, 1 /* send */) = -1 ENOTCONN (Transport endpoint is not connected)
close(2) = 0
There's no need to perform the shutdown() here, the socket is already
in error so it is down.
A check was performed in buffer_replace2() to compare buffer
length with its read pointer. This has been wrong for a long
time, though it only has an impact when dealing with keep-alive
requests/responses. In theory this should be backported but
the check has no impact without keep-alive.
We can receive data with a notification of socket error. But we
must not check for the error before reading the data, because it
may be an asynchronous error notification that we check too early
while the response we're waiting for is available. If there is an
error, recv() will get it.
This should help with servers that close very fast after the response
and should also slightly lower the CPU usage during very fast checks
on massive amounts of servers since we eliminate one system call.
This should probably be backported to 1.3.
While waiting in a keep-alive state for a request, we want to silently
close if we don't get anything. However if we get a partial request it's
different because that means the client has started to send something.
This requires a new transaction flag. It will be used to implement a
distinct timeout for keep-alive and requests.
This change, suggested by Cyril Bont, makes a lot of sense and
would have made it obvious that sessid was not properly initialized
while switching to keep-alive. The code is now cleaner.
The stream_int_cond_close() function was added to preserve the
contents of the response buffer because stream_int_retnclose()
was buggy. It flushed the response instead of flushing the
request. This caused issues with pipelined redirects followed
by error messages which ate the previous response.
This might even have caused object truncation on pipelined
requests followed by an error or by a server redirection.
Now that this is fixed, simply get rid of the now useless
function.
I've tried to follow all the pool_alloc2/pool_free2 calls in the code
to track memory leaks. I've found one which only happens when there's
already no more memory when allocating a new appsession cookie.
Sometimes it can be desired to return a location which is the same
as the request with a slash appended when there was not one in the
request. A typical use of this is for sending a 301 so that people
don't reference links without the trailing slash. The name of the
new option is "append-slash" and it can be used on "redirect"
statements in prefix mode.
When using server redirection, it is possible to specify a path
consisting of only one slash. While this is discouraged (risk of
loop) it may sometimes be useful combined with content switching.
The prefixing of a '/' then causes two slashes to be returned in
the response. So we now do as with the other redirects, don't
prepend a slash if it's alone.
Some message pointers were not usable once the message reached the
HTTP_MSG_DONE state. This is the case for ->som which points to the
body because it is needed to parse chunks. There is one case where
we need the beginning of the message : server redirect. We have to
call http_get_path() after the request has been parsed. So we rely
on ->sol without counting on ->som. In order to achieve this, we're
making ->rq.{u,v} relative to the beginning of the message instead
of the buffer. That simplifies the code and makes it cleaner.
Preliminary tests show this is OK.
This might have been introduced with chunk extensions. Note that
the server redirect still does not work because http_get_path()
cannot get the correct path once the request message is in the
HTTP_MSG_DONE state (->som does not point to the start of message
anymore).
The initial code's intention was to loop on the analysers as long
as an analyser is added by another one. [This code was wrong due to
the while(0) which breaks even on a continue statement, but the
initial intention must be changed too]. In fact we should limit the
number of times we loop on analysers in order to limit latency.
Using maxpollevents as a limit makes sense since this tunable is
used for the exact same purposes. We may add another tunable later
if that ever makes sense, so it's very unlikely.
If we accept a new request and that request produces an immediate
response (error, redirect, ...), then we may fail to send it in
case of pipelined requests if the response buffer is full. To avoid
this, we check the availability of at least maxrewrite bytes in the
response buffer before accepting a new pipelined request.
During a redirect, we used to send the last chunk of response with
stream_int_cond_close(). But this is wrong in case of pipeline,
because if the response already contains something, this function
will refrain from touching the buffer. Use a concatenation function
instead.
Also, this call might still fail when the buffer is full, we need
a second fix to refrain from parsing an HTTP request as long as the
response buffer is full, otherwise we may not even be able to return
a pending redirect or an error code.
That patch was incorrect because under some circumstances, the
capture memory could be freed by session_free() and then again
by http_end_txn(), causing a double free and an eventual segfault.
The pool use count was also reported wrong due to this bug.
The cleanup code was removed from session_free() to remain only
in http_end_txn().
Hank A. Paulson reported a massive memory leak when using keep-alive
mode. The information he provided made it easy to find that captured
request and response headers were erased but not released when renewing
a request.
Several HTTP analysers used to set those flags to values that
were useful but without considering the possibility that they
were not called again to clean what they did. First, replace
direct flag manipulation with more explicit macros. Second,
enforce a rule stating that any buffer which changes one of
these flags from the default must restore it after completion,
so that other analysers see correct flags.
With both this fix and the previous one about analyser bits,
we should not see any more stuck sessions.
A request analyser may very well be added while processing a response
(eg: end of an HTTP keep-alive response). It's very dangerous to only
rely on flags that ought to change in order to loop back, so let's
correctly detect a possible new analyser addition instead of guessing.
With the introduction of keep-alive, we have created situations
where an analyser can add other analysers to the current list,
which are behind it, which have already been processed once, and
which are needed immediately because without them there will be
no more I/O activity. This is typically the case for enabling
reading of a new request after preparing for a new request.
Instead of creating specific cases for some analysers (there was
already one such before), we now use a little bit of algorithmics
to create an ordered bit chain supporting priorities and fast
operations.
Another advantage of this new construction is that it's not a
real loop anymore, so if an analyser is unknown, it will not
loop but just ignore it.
Note that it is easy to skip multiple analysers at once now in
order to speed up the checking a bit. Some test code has shown
a minor gain though.
This change has been carefully re-read and has no direct reason
of causing a regression. However it has been tagged "major"
because the fact that it runs the analysers correctly might
trigger an old sleeping bug somewhere in one of the analysers.
This patch implements default-server support allowing to change
default server options. It can be used in [defaults] or [backend]/[listen]
sections. Currently the following options are supported:
- error-limit
- fall
- inter
- fastinter
- downinter
- maxconn
- maxqueue
- minconn
- on-error
- port
- rise
- slowstart
- weight
Supported informations, available via "tr/td title":
- cap: capabilities (proxy)
- mode: one of tcp, http or health (proxy)
- id: SNMP ID (proxy, socket, server)
- IP (socket, server)
- cookie (backend, server)
This patch adds add "a link" & "a href" html tags for sockets.
As sockets may have the same name like servers, I decided to
add "+" char (forbidden in names assigned to servers), as a prefix.
It is useless to report statistics if the feature was not enabled.
It also makes possible to distinguish if health analyses is
enabled or not only by looking at the stats page.
Commit 0dfdf19b64 introduced a
regression because the connection header is now parsed and checked
depending on the configured options, but the options are set after
calling it instead of being set before.
Historically, "option httpclose" has always worked the same way. It
only mangles the "Connection" header in the request and the response
if needed, but does not affect the connection by itself, and ignores
any further data. It is dangerous to change this behaviour without
leaving any other alternative. If an active close is desired, it's
better to make use of "option forceclose" which does exactly what
it intends to do.
So as of now, "option httpclose" will only mangle the headers as
before, and will only affect the connection by itself when combined
with another connection-related option (eg: keepalive or server-close).
We basically have to mimmic the code of process_session() here, so
when the remote output is closed, we must abort otherwise we'll end
up with data which cannot leave the buffer.
By default this function returned 0 indicating an end of analysis.
This was not a problem as long as it was the last analyser in the
chain but becomes quite a big one now since it skips the forwarder
with auto_close enabled, causing some data to pass under the nose
of the last one undetected.
There were still several situations leading to CLOSE_WAIT sockets
remaining there forever because some complex transitions were
obviously not caught due to the impossibility to resync changes
between the request and response FSMs.
This patch now centralizes the global transaction state and feeds
it from both request and response transitions. That way, whoever
finishes first, there will be no issue for converging to the correct
state.
Some heavy use of the new debugging function has helped a lot. Maybe
those calls could be removed after some time. First tests are very
positive.
This function outputs to fd #-1 the status of request and response
buffers, the transaction states, the stream interface states, etc...
That way, it's easy to find that output in an strace report, correctly
placed WRT the other syscalls.
The data forwarders are analysers. As such, the have to check for
various situations on which they have to abort, one of them being
the lack of data with closed input. Now we don't leave the functions
anymore without performing these checks. This has solved the new
CLOSE_WAIT issue that became more noticeable since last patch.
It may happen that we forward a close just after we sent the last
chunk, because we forgot to clear the AUTO_CLOSE flag.
This issue caused some pages to be truncated depending on some
timing races. Issue initially reported by Cyril Bont.
The cookie parser could be fooled by spaces or commas in cookie names
and values, causing the persistence cookie not to be matched if located
just after such a cookie. Now spaces found in values are considered as
part of the value, and spaces, commas and semi-colons found in values
or names, are skipped till next cookie name.
This fix must be backported to 1.3.
In case of a non-blocking socket, used for connecting to a remote
server (not localhost), the error reported by the health check
was most of a time one of EINPROGRESS/EAGAIN/EALREADY.
This patch adds a getsockopt(..., SO_ERROR, ...) call so now
the proper error message is reported.
It makes sense to permit a client to keep its connection when
performing a redirect to the same host. We only detect the fact
that the redirect location begins with a slash to use the keep-alive
(if the client supports it).
By default we automatically wait for enough data to fill large
packets if buf->to_forward is not null. This causes a problem
with POST/Expect requests which have a data size but no data
immediately available. Instead of causing noticeable delays on
such requests, simply add a flag to disable waiting when sending
requests.
In server-close mode particularly, the response buffer is marked for
no-auto-close after a response passed through. This prevented a POST
request from being aborted on errors, timeouts or anything if the
response was received before the request was complete.
If we enable reading of a request immediately after completing
another one, we end up performing small reads until the request
buffer is complete. This takes time and makes it harder to realign
the buffer when needed. Just enable reading when we need to.
The rq.u field is relative to buf->data, not to msg->sol. We have
to subtract msg->som everywhere this error was made. Maybe it will
be simpler to have a pointer to the buffer in the message and find
appropriate data there.
Many times we see a lot of short responses in HTTP (typically 304 on a
reload). It is a waste of network bandwidth to send that many small packets
when we know we can merge them. When we know that another HTTP request is
following a response, we set BF_EXPECT_MORE on the response buffer, which
will turn MSG_MORE on exactly once. That way, multiple short responses can
leave pipelined if their corresponding requests were also pipelined.
While it could be dangerous to enable MSG_MORE on infinite data (eg:
interactive sessions), it makes sense to enable it when we know the
chunk to be sent is just a part of a larger one.
We used to forward more trailers than required, causing a
desynchronization of the output. Now we schedule all for forwarding
as soon as we encounter them.
This option enables HTTP keep-alive on the client side and close mode
on the server side. This offers the best latency on the slow client
side, and still saves as many resources as possible on the server side
by actively closing connections. Pipelining is supported on both requests
and responses, though there is currently no reason to get pipelined
responses.
When too large a message lies in a buffer before parsing a new
request/response, we can now wait for previous outgoing data to
leave the buffer before attempting to parse again. After that
we can consider the opportunity to realign the buffer if needed.
The HTTP parser needed the msg structure to hold pre-initialized pointers.
This causes a trouble with keep-alive because if some data is still in the
buffer, the pointers can be anywhere after the data and later become invalid
when the buffer gets realigned.
It was not needed to rely on that since we have two valid information
in the buffer itself :
- buf->lr : last visited place
- buf->w + buf->send_max : beginning of next message
So by doing the maths only on those values, we can avoid doing tricks
on msg->som.
This option was disabled for frontends in the configuration because
it was useless in its initial implementation, though it was still
checked in the code. Let's officially enable it now.
When we catch an error from the server, speed up the connection
abort since we don't want to remain long with pending data in the
socket, and we want to be able to reuse our source port ASAP.
Doing this helps us flush the system buffers from all unread data. This
avoids having orphans when clients suddenly get off the net without
reading their entire response.
This new flag may be set by any user on a stream interface to tell
the underlying protocol that there is no need for lingering on the
socket since we know the other side either received everything or
does not care about what we sent.
This will typically be used with forced server close in HTTP mode,
where we want to quickly close a server connection after receiving
its response. Otherwise the system would prevent us from reusing
the same port for some time.
The "forceclose" option used to close the output channel to the
server once it started to respond. While this happened to work with
most servers, some of them considered this as a connection abort and
immediately stopped responding.
Now that we're aware of the end of a request and response, we're able
to trivially handle this option and properly close both sides when the
server's response is complete.
During this change it appeared that forwarding could be allowed when
the BF_SHUTW_NOW flag was set on a buffer, which obviously is not
acceptable and was causing some trouble. This has been fixed too and
is the reason for the MEDIUM status on this patch.
Since we'll soon be able to close a connection with remaining data in a
buffer, it becomes obvious that we can prepare to close when we're about
to send the last chunk of data and not the whole buffer.
This error was triggered by requests not starting at the beginning
of the buffer. It cannot happen with earlier versions though it might
be a good idea to fix it anyway.
There were still issues with the buffer alignment. Now we ensure
that we always align it before a request or response is completely
parsed if there is less than maxrewrite bytes free at the end. In
practice, it's not called that often and ensures we can always work
as expected.
Since the introduction of the automatic sizing of buffers during reads,
a bug appeared where the max size could be negative, causing large
chunks of memory to be overwritten during recv() calls if a read pointer
was already past the buffer's limit.
In many places where we perform header insertion, an error control
is performed but due to a mistake, it cannot match any error :
if (unlikely(error) < 0)
instead of
if (unlikely(error < 0))
This prevents error 400 responses from being sent when the buffer is
full due to many header additions. This must be backported to 1.3.
The body parser will be used in close and keep-alive modes. It follows
the stream to keep in sync with both the request and the response message.
Both chunked transfer-coding and content-length are supported according to
RFC2616.
The multipart/byterange encoding has not yet been implemented and if not
seconded by any of the two other ones, will be forwarded till the close,
as requested by the specification.
Both the request and the response analysers converge into an HTTP_MSG_DONE
state where it will be possible to force a close (option forceclose) or to
restart with a fresh new transaction and maintain keep-alive.
This change is important. All tests are OK but any possible behaviour
change with "option httpclose" might find its root here.
When parsing body for URL parameters, we must not consider that
data are available from buf->data but from buf->data + msg->som.
This is not a problem right now but may become with keep-alive.
When parsing a request that does not start at the beginning of the
buffer, we may experience a buffer full issue. In order to avoid
this, we try to realign the buffer if it is not really full. That
will be required when we have to deal with pipelined requests.
Some wrong operations were performed on buffers, assuming the
offsets were relative to the beginning of the request while they
are relative to the beginning of the buffer. In practice this is
not yet an issue since both are the same... until we add support
for keep-alive.
It's not enough to know if the connection will be in CLOSE or TUNNEL mode,
we still need to know whether we want to read a full message to a known
length or read it till the end just as in TUNNEL mode. Some updates to the
RFC clarify slightly better the corner cases, in particular for the case
where a non-chunked encoding is used last.
Now we also take care of adding a proper "connection: close" to messages
whose size could not be determined.
Chunked encoding can be slightly more complex than what was implemented.
Specifically, it supports some optional extensions that were not parsed
till now if present, and would have caused an error to be returned.
Also, now we enforce check for too large values in chunk sizes in order
to ensure we never overflow.
Last, we're now able to return a request error if we can't read the
chunk size because the buffer is already full.
This state indicates that an HTTP message (request or response) is
complete. This will be used to know when we can re-initialize a
new transaction. Right now we only switch to it after the end of
headers if there is no data. When other analysers are implemented,
we can switch to this state too.
The condition to reuse a connection is when the response finishes
after the request. This will have to be checked when setting the
state.
The response 1xx was set too low and required a lot of tests along
the code in order to avoid some processing. We still left the test
after the response rewrite rules so that we can eliminate unwanted
headers if required.
This code really belongs to the http part since it's transaction-specific.
This will also make it easier to later reinitialize a transaction in order
to support keepalive.
We used to apply a limit to each buffer's size in order to leave
some room to rewrite headers, then we used to remove this limit
once the session switched to a data state.
Proceeding that way becomes a problem with keepalive because we
have to know when to stop reading too much data into the buffer
so that we can leave some room again to process next requests.
The principle we adopt here consists in only relying on to_forward+send_max.
Indeed, both of those data define how many bytes will leave the buffer.
So as long as their sum is larger than maxrewrite, we can safely
fill the buffers. If they are smaller, then we refrain from filling
the buffer. This means that we won't risk to fill buffers when
reading last data chunk followed by a POST request and its contents.
The only impact identified so far is that we must ensure that the
BF_FULL flag is correctly dropped when starting to forward. Right
now this is OK because nobody inflates to_forward without using
buffer_forward().
Up to now, we only had a flag in the session indicating if it had to
work in "connection: close" mode. This is not at all compatible with
keep-alive.
Now we ensure that both sides of a connection act independantly and
only relative to the transaction. The HTTP version of the request
and response is also correctly considered. The connection already
knows several modes :
- tunnel (CONNECT or no option in the config)
- keep-alive (when permitted by configuration)
- server-close (close the server side, not the client)
- close (close both sides)
This change carefully detects all situations to find whether a request
can be fully processed in its mode according to the configuration. Then
the response is also checked and tested to fix corner cases which can
happen with different HTTP versions on both sides (eg: a 1.0 client
asks for explicit keep-alive, and the server responds with 1.1 without
a header).
The mode is selected by a capability elimination algorithm which
automatically focuses on the least capable agent between the client,
the frontend, the backend and the server. This ensures we won't get
undesired situtations where one of the 4 "agents" is not able to
process a transaction.
No "Connection: close" header will be added anymore to HTTP/1.0 requests
or responses since they're already in close mode.
The server-close mode is still not completely implemented. The response
needs to be rewritten as keep-alive before being sent to the client if
the connection was already in server-close (which implies the request
was in keep-alive) and if the response has a content-length or a
transfer-encoding (but only if client supports 1.1).
A later improvement in server-close mode would probably be to detect
some situations where it's interesting to close the response (eg:
redirections with remote locations). But even then, the client might
close by itself.
It's also worth noting that in tunnel mode, no connection header is
affected in either direction. A tunnelled connection should theorically
be notified at the session level, but this is useless since by definition
there will not be any more requests on it. Thus, we don't need to add a
flag into the session right now.
Now that the HTTP analyser will already have parsed the beginning
of the request body, we don't have to check for transfer-encoding
anymore since we have the current chunk size in hdr_content_len.
The POST body analysis was split between two analysers for historical
reasons. Now we only have one analyser which checks content length
and waits for enough data to come.
Right now this analyser waits for <url_param_post_limit> bytes of
body to reach the buffer, or the first chunk. But this could be
improved to wait for any other amount of data or any specific
contents.
The previous check was correct: the RFC states that it is required
to have a domain-name which contained a dot AND began with a dot.
However, currently some (all?) browsers do not obey this specification,
so such configuration might work.
This patch reverts 3d8fbb6658 but
changes the check from FATAL to WARNING and extends the message.
Fix 500b8f0349 fixed the patch for the 64 bit
case but caused the opposite type issue to appear on 32 bit platforms. Cast
the difference and be done with it since gcc does not agree on type carrying
the difference between two pointers on 32 and 64 bit platforms.
Implement decreasing health based on observing communication between
HAProxy and servers.
Changes in this version 2:
- documentation
- close race between a started check and health analysis event
- don't force fastinter if it is not set
- better names for options
- layer4 support
Changes in this version 3:
- add stats
- port to the current 1.4 tree
Cyril Bont found that when an error is detected in one config file, it
is also reported in all other ones, which is wrong. The fix obviously
consists in checking the return code from readcfgfile() and not the
accumulator.
Today I was testing headers manipulation but I met a bug with my first test.
To reproduce it, add for example this line :
rspadd Cache-Control:\ max-age=1500
Check the response header, it will provide :
Cache-Control: max-age=15000 <= the last character is duplicated
This only happens when we use backslashes on the last line of the
configuration file, without returning to the line.
Also if the last line is like :
rspadd Cache-Control:\ max-age=1500\
the last backslash causes a segfault.
This is not due to rspadd but to a more general bug in cfgparse.c :
...
if (skip) {
memmove(line + 1, line + 1 + skip, end - (line + skip + 1));
end -= skip;
}
...
should be :
...
if (skip) {
memmove(line + 1, line + 1 + skip, end - (line + skip));
end -= skip;
}
...
I've reproduced it with haproxy 1.3.22 and the last 1.4 snapshot.
In some environments it is not possible to rely on any wildcard for a
domain name (eg: .com, .net, .fr...) so it is required to send multiple
domain extensions. (Un)fortunately the syntax check on the domain name
prevented that from being done the dirty way. So let's just build a
domain list when multiple domains are passed on the same line.
(cherry picked from commit 950245ca2b)
It was a OR instead of a AND, so it was required to have a cookie
name which contained a dot AND began with a dot.
(cherry picked from commit a1e107fc13)
Gabriel Sosa reported that logs were appearing with BADREQ when
'option httplog' was used with a TCP proxy (eg: inherited via a
default instance). This patch detects it and falls back to tcplog
after emitting a warning.
(cherry picked from commit 5f0bd6537f)
Holger Just reported that running ACLs with too many args caused
a segfault during config parsing. This is caused by a wrong test
on argument count. In case of too many arguments on a config line,
the last one was not correctly zeroed. This is now done and we
report the error indicating what part had been truncated.
(cherry picked from commit 3b39c1446b)
Cameron Simpson reported an annoying case where haproxy simply reports
"Error(s) found in configuration file" when the file is not found or
not readable.
Fortunately the parsing function still returns -1 in case of open
error, so we're able to detect the issue from the caller and report
the corresponding errno message.
In order to support keepalive, we'll have to differentiate
normal sessions from tunnel sessions, which are the ones we
don't want to analyse further.
Those are typically the CONNECT requests where we don't care
about any form of content-length, as well as the requests
which are forwarded on non-close and non-keepalive proxies.
To sum up :
- len : it's now the max number of characters for the value, preventing
garbaged results.
- a new option "prefix" is added, this allows to use dynamic cookie
names (e.g. ASPSESSIONIDXXX).
Previously in the thread, I wanted to use the value found with
"capture cookie" but when i started to update the documentation, I
found this solution quite weird. I've made a small rework to not
depend on "capture cookie".
- There's the posssiblity to define the URL parser mode (path parameters
or query string).
We now set msg->col and msg->sov to the first byte of non-header.
They will be used later when parsing chunks. A new macro was added
to perform size additions on an http_msg in order to limit the risks
of copy-paste in the long term.
During this operation, it appeared that the http_msg struct was not
optimal on 64-bit, so it was re-ordered to fill the holes.
Yohan Tordjman at Dstorage found that upgrading haproxy to 1.4-dev4
caused truncated objects to be returned. An strace quickly exhibited
the issue which was 100% reproducible :
4297 epoll_wait(0, {}, 10, 0) = 0
4297 epoll_wait(0, {{EPOLLIN, {u32=7, u64=7}}}, 10, 1000) = 1
4297 splice(0x7, 0, 0x5, 0, 0xffffffffffffffff, 0x3) = -1 EINVAL (Invalid argument)
4297 shutdown(7, 1 /* send */) = 0
4297 close(7) = 0
4297 shutdown(2, 1 /* send */) = 0
4297 close(2) = 0
This is caused by the fact that the forward length is taken from
BUF_INFINITE_FORWARD, which is -1. The problem does not appear
in 32-bit mode because this value is first cast to an unsigned
long, truncating it to 32-bit (4 GB). Setting an upper bound
fixes the issue.
Also, a second error check has been added for splice. If EINVAL
is returned, we fall back to recv().
An HTTP message can be decomposed into several sub-states depending
on the transfer-encoding. We'll have to keep these state information
while parsing chunks, so we must extend the values. In order not to
change everything, we'll now consider that anything >= MSG_BODY is
the body, and that the value indicates the precise state. The
MSG_ERROR status which was greater than MSG_BODY was moved for this.
Right now, an HTTP server cannot track a TCP server and vice-versa.
This patch enables proxy tracking without relying on the proxy's mode
(tcp/http/health). It only requires a matching proxy name to exist. The
original function was renamed to findproxy_mode().
This patch extends and corrects the functionality introduced by
"Collect & provide http response codes received from servers":
- responses are now also accounted for frontends
- backend's and frontend's counters are incremented based
on responses sent to client, not received from servers
This patch adds <a href> html links for proxies, frontends, servers
and backends. Once located, can be clicked. Users no longer have to
manually add #anchor to stat's url.
All files referencing the previous ebtree code were changed to point
to the new one in the ebtree directory. A makefile variable (EBTREE_DIR)
is also available to use files from another directory.
The ability to build the libebtree library temporarily remains disabled
because it can have an impact on some existing toolchains and does not
appear worth it in the medium term if we add support for multi-criteria
stickiness for instance.
We also check the close status and terminate the server persistent
connection if appropriate. Note that since this change, we'll not
get any "Connection: close" headers added to HTTP/1.0 responses
anymore, which is good.
The code part which waits for an HTTP response has been extracted
from the old function. We now have two analysers and the second one
may re-enable the first one when an 1xx response is encountered.
This has been tested and works.
The calls to stream_int_return() that were remaining in the wait
analyser have been converted to stream_int_retnclose().
Store those elements in the transaction. RFC2616 is strictly followed.
Note that requests containing two different content-length fields are
discarded as invalid.
This patch has 2 goals :
1. I wanted to test the appsession feature with a small PHP code,
using PHPSESSID. The problem is that when PHP gets an unknown session
id, it creates a new one with this ID. So, when sending an unknown
session to PHP, persistance is broken : haproxy won't see any new
cookie in the response and will never attach this session to a
specific server.
This also happens when you restart haproxy : the internal hash becomes
empty and all sessions loose their persistance (load balancing the
requests on all backend servers, creating a new session on each one).
For a user, it's like the service is unusable.
The patch modifies the code to make haproxy also learn the persistance
from the client : if no session is sent from the server, then the
session id found in the client part (using the URI or the client cookie)
is used to associated the server that gave the response.
As it's probably not a feature usable in all cases, I added an option
to enable it (by default it's disabled). The syntax of appsession becomes :
appsession <cookie> len <length> timeout <holdtime> [request-learn]
This helps haproxy repair the persistance (with the risk of losing its
session at the next request, as the user will probably not be load
balanced to the same server the first time).
2. This patch also tries to reduce the memory usage.
Here is a little example to explain the current behaviour :
- Take a Tomcat server where /session.jsp is valid.
- Send a request using a cookie with an unknown value AND a path
parameter with another unknown value :
curl -b "JSESSIONID=12345678901234567890123456789012" http://<haproxy>/session.jsp;jsessionid=00000000000000000000000000000001
(I know, it's unexpected to have a request like that on a live service)
Here, haproxy finds the URI session ID and stores it in its internal
hash (with no server associated). But it also finds the cookie session
ID and stores it again.
- As a result, session.jsp sends a new session ID also stored in the
internal hash, with a server associated.
=> For 1 request, haproxy has stored 3 entries, with only 1 which will be usable
The patch modifies the behaviour to store only 1 entry (maximum).
When processing a GET or HEAD request in close mode, we know we don't
need to read anything anymore on the socket, so we can disable it.
Doing this can save up to 40% of the recv calls, and half of the
epoll_ctl calls.
For this we need a buffer flag indicating that we're not interesting in
reading anymore. Right now, this flag also disables both polled reads.
We might benefit from disabling only speculative reads, but we will need
at least this flag when we want to support keepalive anyway.
Currently we don't disable the flag on completion, but it does not
matter as we close ASAP when performing the shutw().
The fd_list[] used by sepoll was indexed on the fd number and was only
used to store the equivalent of an integer. Changing it to be merged
with fdtab reduces the number of pointer computations, the code size
and some initialization steps. It does not harm other pollers much
either, as only one integer was added to the fdtab array.
Some rarely information are stored in fdtab, making it larger for no
reason (source port ranges, remote address, ...). Such information
lie there because the checks can't find them anywhere else. The goal
will be to move these information to the stream interface once the
checks make use of it.
For now, we move them to an fdinfo array. This simple change might
have improved the cache hit ratio a little bit because a 0.5% of
performance increase has measured.
Till now we would only set SN_CONN_CLOSED after rewriting it. Now we
set it just after checking the Connection header so that we can use
the result later if required.
This can ensure that data is readily available on a socket when
we accept it, but a bug in the kernel ignores the timeout so the
socket can remain pending as long as the client does not talk.
Use with care.
This patch makes stats page about 30% smaller and
"CSS 2.1" + "HTML 4.01 Transitional" compliant.
There should be no visible differences.
Changes:
- add DOCTYPE for HTML 4.01 Transitional
- add missing </ul>
- remove cols=, AFAIK no modern browser support this property and
it prevents validation to pass.
- remove "align: center": there is no such property in css. There is
however "text-align: center" but it is definitely not what we would
like to see here.
- by default align .titre to center
- by default align .td to right
- remove all align=right, no longer necessary
- add class=ac (align center): shorter than "align=center" and use it when
necessary
- remove nowrap from td, instead use "white-space: nowrap" in css
Now stats page passes W3C validators for HTML & CSS. We may consider adding
"validated" icons from www.w3.org. ;)
This alone makes a typical HTML stats dump consume 10% CPU less,
because we avoid doing complex printf calls to drop them later.
Only a few common cases have been checked, those which are very
likely to run for nothing.
It is a bit expensive and complex to use to call buffer_feed()
directly from the request parser, and there are risks that some
output messages are lost in case of buffer full. Since most of
these messages are static, let's have a state dedicated to print
these messages and store them in a specific area shared with the
stats in the session. This both reduces code size and risks of
losing output data.
Krzysztof reported that using names only for get weight/set weight
was not enough because it's still possible to have multiple servers
with the same name (and my test config is one of those). He suggested
to be able to designate them by their unique numeric IDs by prefixing
the ID with a dash.
That way we can have :
set weight #120/#2
as well as
get weight static/srv1 10
Capture & display more data from health checks, like
strerror(errno) for L4 failed checks or a first line
from a response for L7 successes/failed checks.
Non ascii or control characters are masked with
chunk_htmlencode() (html stats) or chunk_asciiencode() (logs).
Add two functions to encode input chunk replacing
non-printable, non ascii or special characters
with:
"&#%u;" - chunk_htmlencode
"<%02X>" - chunk_asciiencode
Above functions should be used when adding strings, received
from possible unsafe sources, to html stats or logs.
int get_backend_server(const char *bk_name, const char *sv_name,
struct proxy **bk, struct server **sv);
This function scans the list of backends and servers to retrieve the first
backend and the first server with the given names, and sets them in both
parameters. It returns zero if either is not found, or non-zero and sets
the ones it did not found to NULL. If a NULL pointer is passed for the
backend, only the pointer to the server will be updated.
The stats socket can now run at 3 different levels :
- user
- operator (default one)
- admin
These levels are used to restrict access to some information
and commands. Only the admin can clear all stats. A user cannot
clear anything nor access sensible data such as sessions or
errors.
The most common use of "clear counters" should be to only clear
max values without affecting cumulated values, for instance,
after an incident. So we change "clear counters" to only clear
max values, and add "clear counters all" to clear all counters.
I noticed that in __eb32_insert , if the tree is empty
(root->b[EB_LEFT] == NULL) , the node.bit is not defined.
However in __task_queue there are checks:
- if (last_timer->node.bit < 0)
- if (task->wq.node.bit < last_timer->node.bit)
which might rely upon an undefined value.
This is how I see it:
1. We insert eb32_node in an empty wait queue tree for a task (called by
process_runnable_tasks() ):
Inserting into empty wait queue &task->wq = 0x72a87c8, last_timer
pointer: (nil)
2. Then, we set the last timer to the same address:
Setting last_timer: (nil) to: 0x72a87c8
3. We get a new task to be inserted in the queue (again called by
process_runnable_tasks()) , before the __task_unlink_wq() is called for
the previous task.
4. At this point, we still have last_timer set to 0x72a87c8 , but since
it was inserted in an empty tree, it doesn't have node.bit and the
values above get dereferenced with undefined value.
The bug has no effect right now because the check for equality is still
made, so the next timer will still be queued at the right place anyway,
without any possible side-effect. But it's a pending bug waiting for a
small change somewhere to strike.
Iliya Polihronov
These ACLs are used to check the number of active connections on the
frontend, backend or in a backend's queue. The avg_queue returns the
average number of queued connections per server, and for this, divides
the total number of queued connections by the number of alive servers.
The dst_conn ACL has been slightly changed to more reflect its name and
original usage, which is to return the number of connections on the
destination address/port (the socket) and not the whole frontend.
Consistent hashing provides some interesting advantages over common
hashing. It avoids full redistribution in case of a server failure,
or when expanding the farm. This has a cost however, the hashing is
far from being perfect, as we associate a server to a request by
searching the server with the closest key in a tree. Since servers
appear multiple times based on their weights, it is recommended to
use weights larger than approximately 10-20 in order to smoothen
the distribution a bit.
In some cases, playing with weights will be the only solution to
make a server appear more often and increase chances of being picked,
so stats are very important with consistent hashing.
In order to indicate the type of hashing, use :
hash-type map-based (default, old one)
hash-type consistent (new one)
Consistent hashing can make sense in a cache farm, in order not
to redistribute everyone when a cache changes state. It could also
probably be used for long sessions such as terminal sessions, though
that has not be attempted yet.
More details on this method of hashing here :
http://www.spiteful.com/2008/03/17/programmers-toolbox-part-3-consistent-hashing/
Commit 404e8ab461 introduced
smart checking for stupid acl typos. However, now haproxy shows
the warning even for valid acls, like this one:
acl Cookie-X-NoAccel hdr_reg(cookie) (^|\ |;)X-NoAccel=1(;|$)
If a frontend does not set 'option socket-stats', a 'clear counters'
on the stats socket could segfault because li->counters is NULL. The
correct fix is to check for NULL before as this is a valid situation.
Recent "struct chunk rework" introduced a NULL pointer dereference
and now haproxy segfaults if auth is required for stats but not found.
The reason is that size_t cannot store negative values, but current
code assumes that "len < 0" == uninitialized.
This patch fixes it.
There are a few remaining max values that need to move to counters.
Also, the counters are more often used than some config information,
so get them closer to the other useful struct members for better cache
efficiency.
Until now it was required that every custom ID was above 1000 in order to
avoid conflicts. Now we have the list of all assigned IDs and can automatically
pick the first unused one. This means that it is perfectly possible to interleave
automatic IDs with persistent IDs and the parser will automatically allocate
unused values starting with 1.
When a name or ID conflict is detected, it is sometimes useful to know
where the other one was declared. Now that we have this information,
report it in error messages.
This patch allows to collect & provide separate statistics for each socket.
It can be very useful if you would like to distinguish between traffic
generate by local and remote users or between different types of remote
clients (peerings, domestic, foreign).
Currently no "Session rate" is supported, but adding it should be possible
if we found it useful.
Doing this, we can remove the last BF_HIJACK user and remove
produce_content(). s->data_source could also be removed but
it is currently used to detect if the stats or a server was
used.
