We store the time stamp of last read in the channel in order to
be able to measure some bit rate and pause lengths. We only use
16 bits which were unused for this. We don't need more, as it
allows us to measure with a millisecond precision for up to 65s.
These ones are only reset during transfers. There is a low but non-null
risk that a first full read causes the previous value to be reused and
immediately to immediately set the CF_STREAMER flag. The impact is only
to increase earlier than expected the SSL record size and to use splice().
This bug was already present in 1.4, so a backport is possible.
Summary:
Track and report last session time on the stats page for each server
in every backend, as well as the backend.
This attempts to address the requirement in the ROADMAP
- add a last activity date for each server (req/resp) that will be
displayed in the stats. It will be useful with soft stop.
The stats page reports this as time elapsed since last session. This
change does not adequately address the requirement for long running
session (websocket, RDP... etc).
By having the stream interface pass the CF_STREAMER flag to the
snd_buf() primitive, we're able to tell the send layer whether
we're sending large chunks or small ones.
We use this information in SSL to adjust the max record dynamically.
This results in small chunks respecting tune.ssl.maxrecord at the
beginning of a transfer or for small transfers, with an automatic
switch to full records if the exchanges last long. This allows the
receiver to parse HTML contents on the fly without having to retrieve
16kB of data, which is even more important with small initcwnd since
the receiver does not need to wait for round trips to start fetching
new objects. However, sending large files still produces large chunks.
For example, with tune.ssl.maxrecord = 2859, we see 5 write(2885)
sent in two segments each and 6 write(16421).
This idea was first proposed on the haproxy mailing list by Ilya Grigorik.
This prevents us from passing other useful info and requires the
upper levels to know these flags. Let's use a new flags category
instead : CO_SFL_*. For now, only MSG_MORE has been remapped.
A new tcp-check rule type: connect.
It allows HAProxy to test applications which stand on multiple ports or
multiple applications load-balanced through the same backend.
Since we support HTTP keep-alive, there is no more reason for staying
in tunnel mode by default. It is confusing for new users and creates
more issues than it solves. Option "http-tunnel" is available to force
to use it if really desired.
Switching to KA by default has implied to change the value of some
option flags and some transaction flags so that value zero (default)
matches keep-alive. That explains why more code has been changed than
expected. Tests have been run on the 25 combinations of frontend and
backend options, plus a few with option http-pretend-keepalive, and
no anomaly was found.
The relation between frontend and backends remains the same. Options
have been updated to take precedence over http-keep-alive which is now
implicit.
All references in the doc to haproxy not supporting keep-alive have
been fixed, and the doc for config options has been updated.
At the very beginning of haproxy, there was "option httpclose" to make
haproxy add a "Connection: close" header in both directions to invite
both sides to agree on closing the connection. It did not work with some
rare products, so "option forceclose" was added to do the same and actively
close the connection. Then client-side keep-alive was supported, so option
http-server-close was introduced. Now we have keep-alive with a fourth
option, not to mention the implicit tunnel mode.
The connection configuration has become a total mess because all the
options above may be combined together, despite almost everyone thinking
they cancel each other, as judging from the common problem reports on the
mailing list. Unfortunately, re-reading the doc shows that it's not clear
at all that options may be combined, and the opposite seems more obvious
since they're compared. The most common issue is options being set in the
defaults section that are not negated in other sections, but are just
combined when the user expects them to be overloaded. The migration to
keep-alive by default will only make things worse.
So let's start to address the first problem. A transaction can only work in
5 modes today :
- tunnel : haproxy doesn't bother with what follows the first req/resp
- passive close : option http-close
- forced close : option forceclose
- server close : option http-server-close with keep-alive on the client side
- keep-alive : option http-keep-alive, end to end
All 16 combination for each section fall into one of these cases. Same for
the 256 combinations resulting from frontend+backend different modes.
With this patch, we're doing something slightly different, which will not
change anything for users with valid configs, and will only change the
behaviour for users with unsafe configs. The principle is that these options
may not combined anymore, and that the latest one always overrides all the
other ones, including those inherited from the defaults section. The "no
option xxx" statement is still supported to cancel one option and fall back
to the default one. It is mainly needed to ignore defaults sections (eg:
force the tunnel mode). The frontend+backend combinations have not changed.
So for examplen the following configuration used to put the connection
into forceclose :
defaults http
mode http
option httpclose
frontend foo.
option http-server-close
=> http-server-close+httpclose = forceclose before this patch! Now
the frontend's config replaces the defaults config and results in
the more expected http-server-close.
All 25 combinations of the 5 modes in (frontend,backend) have been
successfully tested.
In order to prepare for upcoming changes, a new "option http-tunnel" was
added. It currently only voids all other options, and has the lowest
precedence when mixed with another option in another frontend/backend.
If no CA file specified on a server line, the config parser will show an error.
Adds an cmdline option '-dV' to re-set verify 'none' as global default on
servers side (previous behavior).
Also adds 'ssl-server-verify' global statement to set global default to
'none' or 'required'.
WARNING: this changes the default verify mode from "none" to "required" on
the server side, and it *will* break insecure setups.
Till now, we had one flag per stick counter to indicate if it was
tracked in a backend or in a frontend. We just had to add another
flag per stick-counter to indicate if it relies on contents or just
connection. These flags are quite painful to maintain and tend to
easily conflict with other flags if their number is changed.
The correct solution consists in moving the flags to the stkctr struct
itself, but currently this struct is made of 2 pointers, so adding a
new entry there to store only two bits will cause at least 16 more bytes
to be eaten per counter due to alignment issues, and we definitely don't
want to waste tens to hundreds of bytes per session just for things that
most users don't use.
Since we only need to store two bits per counter, an intermediate
solution consists in replacing the entry pointer with a composite
value made of the original entry pointer and the two flags in the
2 unused lower bits. If later a need for other flags arises, we'll
have to store them in the struct.
A few inline functions have been added to abstract the retrieval
and assignment of the pointers and flags, resulting in very few
changes. That way there is no more dependence on the number of
stick-counters and their position in the session flags.
Very often we want to associate one or two flags to a pointer, to
put a type on it or whatever. This patch provides this in standard.h
in the form of a few inline functions which combine a void * pointer
with an int and return an unsigned long called a composite address.
The functions allow to individuall set, retrieve both the pointer and
the flags. This is very similar to what is used in ebtree in fact.
One year ago, commit 5d5b5d8 ("MEDIUM: proto_tcp: add support for tracking
L7 information") brought support for tracking L7 information in tcp-request
content rules. Two years earlier, commit 0a4838c ("[MEDIUM] session-counters:
correctly unbind the counters tracked by the backend") used to flush the
backend counters after processing a request.
While that earliest patch was correct at the time, it became wrong after
the second patch was merged. The code does what it says, but the concept
is flawed. "TCP request content" rules are evaluated for each HTTP request
over a single connection. So if such a rule in the frontend decides to
track any L7 information or to track L4 information when an L7 condition
matches, then it is applied to all requests over the same connection even
if they don't match. This means that a rule such as :
tcp-request content track-sc0 src if { path /index.html }
will count one request for index.html, and another one for each of the
objects present on this page that are fetched over the same connection
which sent the initial matching request.
Worse, it is possible to make the code do stupid things by using multiple
counters:
tcp-request content track-sc0 src if { path /foo }
tcp-request content track-sc1 src if { path /bar }
Just sending two requests first, one with /foo, one with /bar, shows
twice the number of requests for all subsequent requests. Just because
both of them persist after the end of the request.
So the decision to flush backend-tracked counters was not the correct
one. In practice, what is important is to flush countent-based rules
since they are the ones evaluated for each request.
Doing so requires new flags in the session however, to keep track of
which stick-counter was tracked by what ruleset. A later change might
make this easier to maintain over time.
This bug is 1.5-specific, no backport to stable is needed.
show pools
Dump the status of internal memory pools. This is useful to track memory
usage when suspecting a memory leak for example. It does exactly the same
as the SIGQUIT when running in foreground except that it does not flush
the pools.
Just like the previous commit, we sometimes want to limit the rate of
incoming SSL connections. While it can be done for a frontend, it was
not possible for a whole process, which makes sense when multiple
processes are running on a system to server multiple customers.
The new global "maxsslrate" setting is usable to fix a limit on the
session rate going to the SSL frontends. The limits applies before
the SSL handshake and not after, so that it saves the SSL stack from
expensive key computations that would finally be aborted before being
accounted for.
The same setting may be changed at run time on the CLI using
"set rate-limit ssl-session global".
It's sometimes useful to be able to limit the connection rate on a machine
running many haproxy instances (eg: per customer) but it removes the ability
for that machine to defend itself against a DoS. Thus, better also provide a
limit on the session rate, which does not include the connections rejected by
"tcp-request connection" rules. This permits to have much higher limits on
the connection rate without having to raise the session rate limit to insane
values.
The limit can be changed on the CLI using "set rate-limit sessions global",
or in the global section using "maxsessrate".
In addition to previous outputs, we also emit the cumulated number of
connections, the cumulated number of requests, the maximum allowed
SSL connection concurrency, the current number of SSL connections and
the cumulated number of SSL connections. This will help troubleshoot
systems which experience memory shortage due to SSL.
This function is used to compute the new polling state based on
the previous state. All pollers have to do this in their update
loop, so better centralize the logic for it.
Currently, each poll loop handles the polled events the same way,
resulting in a lot of duplicated, complex code. Additionally, epoll
was the only one to handle newly created FDs immediately.
So instead, let's move that code to fd.c in a new function dedicated
to this task : fd_process_polled_events(). All pollers now use this
function.
This is the reimplementation of the "done" action : when we experience
a short read, we're almost certain that we've exhausted the system's
buffers and that we'll meet an EAGAIN if we attempt to read again. If
the FD is not yet polled, the stream interface already takes care of
stopping the speculative read. When the FD is already being polled, we
have two options :
- either we're running from a level-triggered poller, in which case
we'd rather report that we've reached the end so that we don't
speculate over the poller and let it report next time data are
available ;
- or we're running from an edge-triggered poller in which case we
have no choice and have to see the EAGAIN to re-enable events.
At the moment we don't have any edge-triggered poller, so it's desirable
to avoid speculative I/O that we know will fail.
Note that this must not be ported to SSL since SSL hides the real
readiness of the file descriptor.
Thanks to this change, we observe no EAGAIN anymore during keep-alive
transfers, and failed recvfrom() are reduced by half in http-server-close
mode (the client-facing side is always being polled and the second recv
can be avoided). Doing so results in about 5% performance increase in
keep-alive mode. Similarly, we used to have up to about 1.6% of EAGAIN
on accept() (1/maxaccept), and these have completely disappeared under
high loads.
It's easier and safer to rely on conn_xprt_ready() everywhere than to
check the flag itself. It will also simplify adding extra checks later
if needed. Some useless controls for !xprt have been removed, as the
XPRT_READY flag itself guarantees xprt is set.
It's easier and safer to rely on conn_ctrl_ready() everywhere than to
check the flag itself. It will also simplify adding extra checks later
if needed. Some useless controls for !ctrl have been removed, as the
CTRL_READY flag itself guarantees ctrl is set.
We simply remove these functions and replace their calls with the
appropriate ones :
- if we're in the data phase, we can simply report wait on the FD
- if we're in the socket phase, we may also have to signal the
desire to read/write on the socket because it might not be
active yet.
These flags were used to report the readiness of the file descriptor.
Now this readiness is directly checked at the file descriptor itself.
This removes the need for constantly synchronizing updates between the
file descriptor and the connection and ensures that all layers share
the same level of information.
For now, the readiness is updated in conn_{sock,data}_poll_* by directly
touching the file descriptor. This must move to the lower layers instead
so that these functions can disappear as well. In this state, the change
works but is incomplete. It's sensible enough to avoid making it more
complex.
Now the sock/data updates become much simpler because they just have to
enable/disable access to a file descriptor and not to care anymore about
its readiness.
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
We're completely changing the way FDs will be polled. There will be no
more speculative I/O since we'll know the exact FD state, so these will
only be cached events.
First, let's fix a few field names which become confusing. "spec_e" was
used to store a speculative I/O event state. Now we'll store the whole
R/W states for the FD there. "spec_p" was used to store a speculative
I/O cache position. Now let's clearly call it "cache".
We're completely changing the way FDs will be polled. First, let's fix
a few field names which become confusing. "spec_e" was used to store a
speculative I/O event state. Now we'll store the whole R/W states for
the FD there.
Patrick Hemmer reported that using unique_id_format and logs did not
report the same unique ID counter since commit 9f09521 ("BUG/MEDIUM:
unique_id: HTTP request counter must be unique!"). This is because
the increment was done while producing the log message, so it was
performed twice.
A better solution consists in fetching a new value once per request
and saving it in the request or session context for all of this
request's life.
It happens that sessions already have a unique ID field which is used
for debugging and reporting errors, and which differs from the one
sent in logs and unique_id header.
So let's change this to reuse this field to have coherent IDs everywhere.
As of now, a session gets a new unique ID once it is instanciated. This
means that TCP sessions will also benefit from a unique ID that can be
logged. And this ID is renewed for each extra HTTP request received on
an existing session. Thus, all TCP sessions and HTTP requests will have
distinct IDs that will be stable along all their life, and coherent
between all places where they're used (logs, unique_id header,
"show sess", "show errors").
This feature is 1.5-specific, no backport to 1.4 is needed.
It is quite often that an connection error only reports "socket error" with
no more information. This is especially problematic with health checks where
many causes are possible, including resource exhaustion which do not lead to
a valid errno code. So let's add explicit codes to cover these cases.
Till now there was no way to know from a connection if a previous
call to drain() had done any change. This function is used to drain
incoming data and to update the connection's flags at the same time.
It also correctly sets the polling flags on the connection if the
drain function indicates inability to receive. This function will
be used preferably over ctrl->drain() when a connection is used.
Recent commit 4448925 ("BUILD/MINOR: listener: remove a glibc warning on accept4()")
broke accept4() on some systems because the glibc's version may now conflict with
the local one.
This reverts commit 12082663561aa2189d243328060c399f2fd95860.
It randomly breaks SSL. What happens is that if the SSL response is
read at once by the SSL stack and is partially delivered to the buffer,
then there's no way to read the next parts because we wait for some
polling first.
So we'll fix this after the polling rework.
This function is called twice per request, and does almost always nothing.
Better use an inline version to avoid entering it when we can.
About 0.5% additional performance was gained this way.
si_connect() used to only return SI_ST_CON. But it already detect the
connection reuse and is the function which avoids calling connect().
So it already knows the connection is valid and reuse. Thus we make it
return SI_ST_EST when a connection is reused. This means that
connect_server() can return this state and sess_update_stream_int()
as well.
Thanks to this change, we don't need to leave process_session() in
SI_ST_CON state to immediately enter it again to switch to SI_ST_EST.
Implementing this removes one call to process_session() per request
in keep-alive mode. We're now at 2 calls per request, which is the
minimum (one for the request and another one for the response). The
number of calls to http_wait_for_response() has also dropped from 2
to one.
Tests indicate a performance gain of about 2.6% in request rate in
keep-alive mode. There should be no gain in http-server-close() since
we don't use this faster path.
Since commit 6b66f3e ([MAJOR] implement autonomous inter-socket forwarding)
introduced in 1.3.16-rc1, we've been relying on a stupid mechanism to wake
up the task after a write, which was an exact copy-paste of the reader side.
The principle was that if we empty a buffer and there's no forwarding
scheduled or if the *producer* is not in a connected state, then we wake
the task up.
That does not make any sense. It happens to wake up too late sometimes (eg,
when the request analyser waits for some room in the buffer to start to
work), and leads to unneeded wakeups in client-side keep-alive, because
the task is woken up when the response is sent, while the analysers are
simply waiting for a new request.
In order to fix this, we introduce a new channel flag : CF_WAKE_WRITE. It
is designed so that an analyser can explicitly request being notified when
some data were written. It is used only when the HTTP request or response
analysers need to wait for more room in the buffers. It is automatically
cleared upon wake up.
The flag is also automatically set by the functions which try to write into
a buffer from an applet when they fail (bi_putblk() etc...).
That allows us to remove the stupid condition above and avoid some wakeups.
In http-server-close and in http-keep-alive modes, this reduces from 4 to 3
the average number of wakeups per request, and increases the overall
performance by about 1.5%.
This reverts commit f3221f99acdd792352d4ee648d987270d74ca38e.
Igor reported some very strange breakage of his stats page which is
clearly caused by the chunking, though I don't see at first glance
what could be wrong. Better revert it for now.
In theory the principle is simple as we just need to send HTTP chunks
if the client is 1.1 compatible. In practice it's harder because we
have to append a CR LF after each block of data and we're never sure
to have the room for this. In order not to have to deal with this, we
instead send the CR LF prior to each chunk size. The only issue is for
the first chunk and for this reason we avoid to send the empty header
line when using chunked encoding.
This value is stored as unsigned in chn->to_forward. Having it defined
as signed makes it impossible to pass channel_forward() a previously
saved value because the argument will be zero-extended during the
conversion to long long, while the test will be performed using sign
extension. There is no impact on existing code right now.
If a file descriptor is being polled, and it stopped (eg: buffer full
or end of response), then re-enabled, currently what happens is that
the polling is disabled, then the fd is enabled in speculative mode,
an I/O attempt is made, it loses (otherwise the FD would surely not
have been polled), and the polled is enabled again.
This is too bad, especially with HTTP keep-alive on the server side
where all operations are performed at once before going back to the
poll loop.
Now we improve the behaviour by ensuring that if an fd is still being
polled, when it's enabled after having been disabled, we re-enable the
polling. Doing so saves a number of syscalls and useless wakeups, and
results in a significant performance gain on HTTP keep-alive. A 11%
increase has been observed on the HTTP request rate in keep-alive
thanks to this.
It could be considered as a bug fix, but there was no harm with the
current behaviour, except extra syscalls.
In HTTP keep-alive mode, if we receive a 401, we still have a chance
of being able to send the visitor again to the same server over the
same connection. This is required by some broken protocols such as
NTLM, and anyway whenever there is an opportunity for sending the
challenge to the proper place, it's better to do it (at least it
helps with debugging).
Idle connections are not monitored right now. So if a server closes after
a response without advertising it, it won't be detected until a next
request wants to use the connection. This is a bit problematic because
it unnecessarily maintains file descriptors and sockets in an idle
state.
This patch implements a very simple idle connection manager for the stream
interface. It presents itself as an I/O callback. The HTTP engine enables
it when it recycles a connection. If a close or an error is detected on the
underlying socket, it tries to drain as much data as possible from the socket,
detect the close and responds with a close as well, then detaches from the
stream interface.
The throttling of low weight servers (<16) could mistakenly be reported
as > 100% due to a rounding that was performed before a multiply by 100
instead of after. This was introduced in 1.5-dev20 when fixing a previous
reporting issue by commit d32c399 (MINOR: stats: report correct throttling
percentage for servers in slowstart).
It should be backported if the patch above is backported.
When the load balancing algorithm in use is not deterministic, and a previous
request was sent to a server to which haproxy still holds a connection, it is
sometimes desirable that subsequent requests on a same session go to the same
server as much as possible. Note that this is different from persistence, as
we only indicate a preference which haproxy tries to apply without any form
of warranty. The real use is for keep-alive connections sent to servers. When
this option is used, haproxy will try to reuse the same connection that is
attached to the server instead of rebalancing to another server, causing a
close of the connection. This can make sense for static file servers. It does
not make much sense to use this in combination with hashing algorithms.
This is the best place to reuse a connection. We centralize all
connection requests and we're at the best place to know exactly
what the current state of the underlying connection is. If the
connection is reused, we just enable polling for send() in order
to be able to emit the request.
When allocating a new connection, only the caller knows whether it's
acceptable to reuse the previous one or not. Let's pass this information
to si_alloc_conn() which will do the cleanup if the connection is not
acceptable.
