If a CRI error occurs during the terminating phase after a pod is
force deleted (API or static) then the housekeeping loop will not
deliver updates to the pod worker which prevents the pod's state
machine from progressing. The pod will remain in the terminating
phase but no further attempts to terminate or cleanup will occur
until the kubelet is restarted.
The pod worker now maintains a store of the pods state that it is
attempting to reconcile and uses that to resync unknown pods when
SyncKnownPods() is invoked, so that failures in sync methods for
unknown pods no longer hang forever.
The pod worker's store tracks desired updates and the last update
applied on podSyncStatuses. Each goroutine now synchronizes to
acquire the next work item, context, and whether the pod can start.
This synchronization moves the pending update to the stored last
update, which will ensure third parties accessing pod worker state
don't see updates before the pod worker begins synchronizing them.
As a consequence, the update channel becomes a simple notifier
(struct{}) so that SyncKnownPods can coordinate with the pod worker
to create a synthetic pending update for unknown pods (i.e. no one
besides the pod worker has data about those pods). Otherwise the
pending update info would be hidden inside the channel.
In order to properly track pending updates, we have to be very
careful not to mix RunningPods (which are calculated from the
container runtime and are missing all spec info) and config-
sourced pods. Update the pod worker to avoid using ToAPIPod()
and instead require the pod worker to directly use
update.Options.Pod or update.Options.RunningPod for the
correct methods. Add a new SyncTerminatingRuntimePod to prevent
accidental invocations of runtime only pod data.
Finally, fix SyncKnownPods to replay the last valid update for
undesired pods which drives the pod state machine towards
termination, and alter HandlePodCleanups to:
- terminate runtime pods that aren't known to the pod worker
- launch admitted pods that aren't known to the pod worker
Any started pods receive a replay until they reach the finished
state, and then are removed from the pod worker. When a desired
pod is detected as not being in the worker, the usual cause is
that the pod was deleted and recreated with the same UID (almost
always a static pod since API UID reuse is statistically
unlikely). This simplifies the previous restartable pod support.
We are careful to filter for active pods (those not already
terminal or those which have been previously rejected by
admission). We also force a refresh of the runtime cache to
ensure we don't see an older version of the state.
Future changes will allow other components that need to view the
pod worker's actual state (not the desired state the podManager
represents) to retrieve that info from the pod worker.
Several bugs in pod lifecycle have been undetectable at runtime
because the kubelet does not clearly describe the number of pods
in use. To better report, add the following metrics:
kubelet_desired_pods: Pods the pod manager sees
kubelet_active_pods: "Admitted" pods that gate new pods
kubelet_mirror_pods: Mirror pods the kubelet is tracking
kubelet_working_pods: Breakdown of pods from the last sync in
each phase, orphaned state, and static or not
kubelet_restarted_pods_total: A counter for pods that saw a
CREATE before the previous pod with the same UID was finished
kubelet_orphaned_runtime_pods_total: A counter for pods detected
at runtime that were not known to the kubelet. Will be
populated at Kubelet startup and should never be incremented
after.
Add a metric check to our e2e tests that verifies the values are
captured correctly during a serial test, and then verify them in
detail in unit tests.
Adds 23 series to the kubelet /metrics endpoint.
Add some additional init container tests that work via monitoring
container lifetime based on logs written to a common file. This allows
more easily writing assertions about the container lifetimes with
respect to one another.
6f2cd1b5bd swapped the order of cancel() and
closeFn() so that closeFn got called first when the test was done. This caused
it to block while waiting for goroutines which themselves were waiting for
the context cancellation. The test still shut down, it just took ~86s instead
of ~30s.
The fix is to register the cancel twice: once as soon as the context is
created (to clean up in case of an unexpected panic) and once after
closeFn (because then it'll get called first, as before).
The test creates a Service exposing two protocols on the same port
and a backend that replies on both protocols.
1. Test that Service with works for both protocol
2. Update Service to expose only the TCP port
3. Verify that TCP works and UDP does not work
4. Update Service to expose only the UDP port
5. Verify that TCP does not work and UDP does work
Change-Id: Ic4f3a6509e332aa5694d20dfc3b223d7063a7871
Test 2 scenarios:
- pod can connect to a terminating pods
- terminating pod can connect to other pods
Change-Id: Ia5dc4e7370cc055df452bf7cbaddd9901b4d229d
v1.Container is still changing a log which caused the test to fail each time a
new field was added. To test loading, let's better use something that is
unlikely to change. The runtimev1.VersionResponse gets logged by kubelet and
seems to be stable.
The benchmarks and unit tests were written so that they used custom APIs for
each log format. This made them less realistic because there were subtle
differences between the benchmark and a real Kubernetes component. Now all
logging configuration is done with the official
k8s.io/component-base/logs/api/v1.
To make the different test cases more comparable, "messages/s" is now reported
instead of the generic "ns/op".
When trying again with recent log files from the CI job, it was found that some
JSON messages get split across multiple lines, both in container logs and in
the systemd journal:
2022-12-21T07:09:47.914739996Z stderr F {"ts":1671606587914.691,"caller":"rest/request.go:1169","msg":"Response ...
2022-12-21T07:09:47.914984628Z stderr F 70 72 6f 78 79 10 01 1a 13 53 ... \".|\n","v":8}
Note the different time stamp on the second line. That first line is
long (17384 bytes). This seems to happen because the data must pass through a
stream-oriented pipe and thus may get split up by the Linux kernel.
The implication is that lines must get merged whenever the JSON decoder
encounters an incomplete line. The benchmark loader now supports that. To
simplifies this, stripping the non-JSON line prefixes must be done before using
a log as test data.
The updated README explains how to do that when downloading a CI job
result. The amount of manual work gets reduced by committing symlinks under
data to the expected location under ci-kubernetes-kind-e2e-json-logging and
ignoring them when the data is not there.
Support for symlinks gets removed and path/filepath is used instead of path
because it has better Windows support.
A Service can use multiple EndpointSlices for its backend, when
using custom Endpoint Slices, the data plane should forward traffic
to any of the endpoints in the Endpointslices that belong to the
Service.
Change-Id: I80b42522bf6ab443050697a29b94d8245943526f
