The problem was that some of the health checks sort the list of the
nodes in place (via `sort.Strings()`). If cluster info provider returns
original slice, it might be mutated in such a way that it gets
corrupted.
We never noticed it before CAPI clusters, as in our tests IPs are
assigned sequentially, and sort operation is a no-op.
Specifically, the problem was with the `Nodes()` function, it returns
`append(controlPlaneNodes, workerNodes...)` slice, which by definition
might share memory with `controlPlaneNodes` slice. For example,
if control plane nodes were `4, 5, 6` and worker nodes were `3`, the
returned slice will be `4, 5, 6, 3`, and it shares memory with
`controlPlaneNodes` slice (firs three items). If we apply `sort` to the
returned slice, it re-orders it as `3, 4, 5, 6`, but as it is done
in-place, the `controlPlaneNodes` slice is now `3, 4, 5`, which is
obviously wrong.
Fix that by always returning a copy of the slice from the functions
implementing `ClusterInfo` interface.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
It only gets enabled if output is a terminal. Failures which resolve
themselves are removed from the final output. Small spinner to indicate
progress.
While I was at it, I fixed client-side `talosctl health` when init node
is missing.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Kubeconfig merge was completely rewritten to be "smarter":
* automatically apply renames done at previous stages to avoid asking
over and over again (in general should ask just once)
* skip checks if parts of the config match exactly
* allow overwrite as an option
* flexible way to control the output
* activating context in the end
* custom merged context name
Fixes#2578Fixes#2587Fixes#2577
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This also refactors much of the CLI code for the `talosctl kubeconfig`:
1. Do all the checks before fetching kubeconfig from the server: as
kubeconfig generation takes a few seconds, it doesn't make sense to
generate it if it's not going to be used.
2. Unify most of merge & write directly features.
3. Don't use ExtractTarGz method to be more flexible.
4. Allow custom paths for kubeconfig, whether it is a directory or full
path to the file to be created.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Adds the ability to apply (replace) an existing node configuration with
a new one via the Machine API.
Fixes#2345
Signed-off-by: Seán C McCord <ulexus@gmail.com>
By default, build outside of Drone works the same and builds only amd64
version, loads images back into dockerd, etc.
If multiple platforms are used, multi-arch images are built which can't
be exported to docker or to `.tar` image, they're always pushed to the
registry (even for PR builds to our internal CI registry).
Artifacts as files (initramfs, kernel) now have `-arch` suffix:
`vmlinuz-amd64`, `initramfs-amd64.xz`. "Magic" script normalizes output
paths depending on whether single platform or multiple platforms were
given.
VM provisioners accept magic `${ARCH}` in initramfs/kernel paths which
gets replaced by cluster architecture.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This adds a command that lists all of the images used by Talos. This is
useful in the case of airgap installs, so that users will know which images
to pull.
Signed-off-by: Andrew Rynhard <andrew@rynhard.io>
Add Kubernetes upgrade as part of the provisioning (upgrade tests):
first K8s control plane is upgraded, then Talos is upgraded (with
kubelet), and e2e test is run last.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Add sonobuoy runner code with log fetching on failure. Use hand-picked
set of e2e tests to run: verify basic pod functionality, verify service
connectivity.
Add option `--run-e2e` to the `talosctl health` to run quick e2e test to
verify cluster health.
Add option to run provision tests with custom CNI, run one track of
provision tests with Cilium.
Bump Cilium to 1.8.2.
Talos 0.6 won't uncordon node automatically after upgrade from 0.5, as
0.5 doesn't put annotation. Workaround that in upgrade tests.
Bump upgrade test version to 0.6.0 release.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This moves to using grub instead of syslinux.
BREAKING CHANGE: Single node upgrades will fail in this change. This
will also break the A/B fallback setup since this version introduces
an entirely new partition scheme, that any fallback will not know about.
We plan on addressing these issues in a follow up change.
Signed-off-by: Andrew Rynhard <andrew@rynhard.io>
Bootkube recover process (and `talosctl recover`) was actually
regenerating assets each time `recover` runs forcing control plane to be
at the state when cluster got created. This PR fixes that by running
recover process correctly.
Recovery via etcd was fixed to handle encrypted etcd data:
it follows the way `apiserver` handles encryption at rest, and as at
the moment AES CBC is the only supported encryption method, code simply
follows the same path.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This moves `pkg/config`, `pkg/client` and `pkg/constants`
under `pkg/machinery` umbrella.
And `pkg/machinery` is published as Go module inside Talos repository.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This change is only moving packages and updating import paths.
Goal: expose `internal/pkg/provision` as `pkg/provision` to enable other
projects to import Talos provisioning library.
As cluster checks are almost always required as part of provisioning
process, package `internal/pkg/cluster` was also made public as
`pkg/cluster`.
Other changes were direct dependencies discovered by `importvet` which
were updated.
Public packages (useful, general purpose packages with stable API):
* `internal/pkg/conditions` -> `pkg/conditions`
* `internal/pkg/tail` -> `pkg/tail`
Private packages (used only on provisioning library internally):
* `internal/pkg/inmemhttp` -> `pkg/provision/internal/inmemhttp`
* `internal/pkg/kernel/vmlinuz` -> `pkg/provision/internal/vmlinuz`
* `internal/pkg/cniutils` -> `pkg/provision/internal/cniutils`
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Fixes#2363#2364#2370#2371
Several changes packed together:
* use compressed `vmlinuz` everywhere, firecracker provisioner
uncompresses it before first use, drop `vmlinux`
* handle reboots in qemu launcher to support reset API case, update
empty disk check to handle reset behavior (erasing partition table)
* make bootloader support default in provisioners, and flag to disable
that
* early support for target architecture for qemu provisioner
This should allow us to use `qemu` in CI/CD (not included into this PR):
integration test passes with qemu.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This makes `pkg/config` directly importable from other projects.
There should be no functional changes.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This bumps next version to the latest 0.6 alpha and latest 0.5.
This also enables single node preserve test.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Fixes#2330
CLI tests require node discovery as `--nodes` flag is enforced for most
of the `talosctl commands`.
For clusters created via `talosctl cluster create`, cluster provisioner
state provides all the necessary information, but clusters created via
CAPI don't have the state attached.
API tests rely on Talos and Kubernetes APIs to fetch kubeconfig and
access Nodes K8s API.
CLI tests should rely only on CLI tools, so we use `kubectl get nodes` +
`talosctl kubeconfig` to fetch list of master and worker nodes.
This discovery method relies on "bootstrap" node being set in
`talosconfig` (to fetch `kubeconfig`).
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
With load-balancing enabled by default running `talosctl` without
`--nodes` is risky, as it might hit any control plane by default without
`--nodes`.
Only two commands do not enforce this check, as they do their own node
contexts: `crashdump` and `health` (client-side).
Integration tests were updated to always supply `--nodes` cli argument,
while doing that I refactored the storage for discovered nodes to use
existing `cluster.Info` interface.
The downside is that with e2e CAPI tests CLI tests will be mostly
skipped as we don't support discovery in CLI tests at the momemnt. This
can be fixed by using `talosctl kubeconfig` + `kubectl get nodes` for
node discovery.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Fixes#2316
Simply update dependencies we don't track on version level to be
compatible with Talos components (like etcd or k8s).
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This implements existing server-side health checks as defined in
`internal/pkg/cluster/checks` in Talos API.
Summary of changes:
* new `cluster` API
* `apid` now listens without auth on local file socket
* `cluster` API is for now implemented in `machined`, but we can move it
to the new service if we find it more appropriate
* `talosctl health` by default now does server-side health check
UX: `talosctl health` without arguments does health check for the
cluster if it has healthy K8s to return master/worker nodes. If needed,
node list can be overridden with flags.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This merges `osd` API into `machined`. API was copied from `osd` into
`machined`, and `osd` API was deprecated.
For backwards compatibility, `machined` still implements `osd` API, so
older Talos API clients can still talk to the node without changes.
Docs were updated. No functional changes.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Previously we assumed that node 0 is the init node, and it can't be
reset. With new bootstrap API approach, there's no init node, and all
the nodes can be reset. This corrects the check to skip only the init
node, and with bootstrap API there's no init node (so no nodes are
skipped).
Fixes#2277
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Talos will mark node as schedulable if it was previously cordoned by
Talos (for upgrade, reset, etc.)
If user marked node as not schedulable, Talos won't change it on boot.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
There were three problems:
* cli tests did commands in sequence assuming they all hit the same
node, but with load-balancing it's no longer true
* restart test was affected, as it hit different node for check after
restart, and it succeeded immediately, while on original node process
was still starting which resulted in failure in the next tests; replace
the check to make sure service is up and healthy, so that test leaves
cluster in a good state
* restart API response had wrong format (no message returned) which
resulted in failures with apid proxy (when used with `-n`)
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Handling of multiple endpoints has already been implemented in #2094.
This PR enables round-robin policy so that grpc picks up new endpoint
for each call (and not send each request to the first control plane
node).
Endpoint list is randomized to handle cases when only one request is
going to be sent, so that it doesn't go always to the first node in the
list.
gprc handles dead/unresponsive nodes automatically for us.
`talosctl cluster create` and provision tests switched to use
client-side load balancer for Talos API.
On the additional improvements we got:
* `talosctl` now reports correct node IP when using commands without
`-n`, not the loadbalancer IP (if using multiple endpoints of course)
* loadbalancer can't provide reliable handling of errors when upstream
server is unresponsive or there're no upstreams available, grpc returns
much more helpful errors
Fixes#1641
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
1. Add [xid-based](https://github.com/rs/xid) event IDs. Xids
are sortable and unique enough. Xids also encode event publishing
time with a second precision.
2. Add three ways to look back into event history: based on number of
events, on time and ID. Lookup via ID might be used to restart event
polling in case of broken API connection from the same moment.
3. Reimplement core event buffer with positions which are always
incremented instead of generation+index, this implementation is much
more simple (idea from circular buffer).
4. By default, Events API works the same - it shows no history and
starts streaming new events only.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This implements service events, adds test for events API based on
service events as they're the easiest to generate on demand.
Disabled validate test for 'metal' as it validates disk device against
local system which doesn't make much sense.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
These tests rely on node uptime checks. These checks are quite flaky.
Following fixes were applied:
* code was refactored as common method shared between reset/reboot tests
(reboot all nodes does checks in a different way, so it wasn't updated)
* each request to read uptime times out in 5 seconds, so that checks
don't wait forever when node is down (or connection is aborted)
* to account for node availability vs. lower uptime in the beginning of
test, add extra elapsed time to the check condition
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This replaces logging to files with inotify following to pure in-memory
circular buffer which grows on demand capped at specified maximum
capacity.
The concern with previous approach was that logs on tmpfs were growing
without any bound potentially consuming all the node memory.
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
This moves our test scripts to using the bootstrap API. Some
automation around invoking the bootstrap API was also added
to give the same ease of use when creating clusters with the
CLI.
Signed-off-by: Andrew Rynhard <andrew@andrewrynhard.com>
After node reboot test code tries endlessly to read the uptime
until it goes down after reboot, but during actual reboot API won't be
responsive and it might happen that this call will time out only with
parent context canceling, and by that time retry timeout is already
exhausted, so no more attempts will be made (while node successfully
booted after a reboot).
```
uptime didn't go down: before 219.730000, after 267.020000
uptime didn't go down: before 219.730000, after 268.030000
EOF
rpc error: code = DeadlineExceeded desc = context deadline exceeded
timeout
```
Signed-off-by: Andrey Smirnov <smirnov.andrey@gmail.com>
Create a new `client.New` to make external API systems easier to
construct.
A new type `client.OptionFunc` allows the client to be
extended with specific configuration.
This also makes a first pass at supporting multiple endpoints properly
by creating a custom grpc resolver.
(Proper load balancing support is still a TODO.)
Fixes#2093
Signed-off-by: Seán C McCord <ulexus@gmail.com>
