994 Commits

Author SHA1 Message Date
Jan Chaloupka
f4718bf928 refactor(prom client controllers): change the one letter receiver into ctrl 2026-02-05 15:54:49 +01:00
Jan Chaloupka
f149f5a083 refactor: move prometheus client controller related code under a seperate file 2026-02-05 15:42:47 +01:00
Jan Chaloupka
b214c14793 feat(pkg/descheduler): create profiles outside the descheduling cycle 2026-02-04 20:04:31 +01:00
Jan Chaloupka
d262c7af44 refactor(TestPromClientControllerSync_EventHandler): be more verbose about the target expectations 2026-02-04 17:44:17 +01:00
Jan Chaloupka
4b5be0a772 feat(prometheus client reconciling): be more strict about clearing the previous connection
To avoid stalling connections that are not expected to be kept. E.g.
when an invalid secret is provided.
2026-02-04 17:16:44 +01:00
Jan Chaloupka
29e5a51cb5 refactor(newSecretBasedPromClientController): inline setupPrometheusProvider into newSecretBasedPromClientController 2026-02-04 17:16:15 +01:00
Jan Chaloupka
a91a02cadc refactor(newSecretBasedPromClientController): move prometheus config validation under newSecretBasedPromClientController 2026-02-04 17:14:58 +01:00
Jan Chaloupka
964df4ce95 refactor(promClientController): split it into two prom client controllers 2026-02-04 16:59:13 +01:00
Jan Chaloupka
c357bc3d28 test(TestReconcileInClusterSAToken): have prometheus client tests build an entire descheduler as well 2026-02-04 13:26:00 +01:00
Jan Chaloupka
f8a4efd4f0 format(TestReconcileInClusterSAToken): indent the code by two tabs 2026-02-04 13:18:53 +01:00
Jan Chaloupka
0822f6f99a test(TestPromClientControllerSync_EventHandler): have prometheus client tests build an entire descheduler as well 2026-02-04 13:18:07 +01:00
Jan Chaloupka
4361c133e9 format(TestPromClientControllerSync_EventHandler): indent the code by two tabs 2026-02-04 13:11:20 +01:00
Jan Chaloupka
818675f73e refactor(TestPromClientControllerSync_EventHandler): get secret name and namespace from the implicitly generated secret object 2026-02-04 12:47:41 +01:00
Jan Chaloupka
a8f618c3fd test(TestPromClientControllerSync_ClientCreation): have prometheus client tests build an entire descheduler as well 2026-02-04 12:41:31 +01:00
Jan Chaloupka
760bacab34 format(TestPromClientControllerSync_ClientCreation): indent the code by two tabs 2026-02-04 12:40:12 +01:00
Jan Chaloupka
aa21eed809 refactor(setupPromClientControllerTest): inject context instead of creating a stop channel 2026-02-04 12:33:53 +01:00
Jan Chaloupka
807f580204 test(pkg/descheduler): TestPluginPrometheusClientAccess_Secret to test secret based token reconciliation
This test makes sure the secret is reconciled and available in plugins
while running the descheduler from its main entry point.
2026-02-04 11:31:18 +01:00
Jan Chaloupka
643179cdac feat(pkg/descheduler): setup prometheus client for the dry run mode descheduler as well
The secret is used for read-only operations so it can be safely wired.
2026-02-04 11:29:09 +01:00
Jan Chaloupka
d1671034df refactor(pkg/descheduler): move the main descheduler run entry under a dedicated function 2026-02-03 23:31:29 +01:00
Jan Chaloupka
f26ab740fe feat(pkg/descheduler): do not cancel the context when runDeschedulerLoop fails
When the descheduler is running in the dry mode the the kube client sandbox restoring may fail.
Which can be caused by timeouts when waiting for internal caches to sync.
The internal timeouts depend on the cluster size which changes in time.
No reason to cancel the context because of that.
2026-02-03 23:31:06 +01:00
Jan Chaloupka
0d05e5dd1f refactor(pkg/descheduler): dedicated function for setting token reconciliation 2026-02-03 23:31:01 +01:00
Jan Chaloupka
e402adb9f1 refactor(pkg/descheduler): start factory informers and other descheduler parts during bootstrap
This will help with testing as the initDescheduler function under
descheduler_test.go is now much closer to the production invocation.
2026-02-03 21:33:28 +01:00
Jan Chaloupka
0c307cf7b9 refactor(pkg/descheduler): bootstrap a descheduler through a dedicated function 2026-02-03 21:26:16 +01:00
Jan Chaloupka
a71c2ab331 refactor(pkg/descheduler): single ctx with cancel with the whole RunDeschedulerStrategies function for consistency 2026-02-03 21:15:13 +01:00
Kubernetes Prow Robot
ffd05ff9c1 Merge pull request #1818 from ingvagabund/prom-client-testing
tests: Prom client testing
2026-02-02 19:04:28 +05:30
Kubernetes Prow Robot
a60ab7b5ed Merge pull request #1786 from W1seKappa/add-namespace-label-selector
Add namespace label selector
2026-02-02 18:24:29 +05:30
Danila Bobkov
4361eefaf2 add namespaceLabelSelector
Signed-off-by: Danila Bobkov <danila.bobkov@flant.com>
2026-02-02 15:22:11 +03:00
Jan Chaloupka
5a53f16526 test(promClientController): unit test the implementation
Currently, there's a single prometheus client reconciler for both in
cluster and secret based strategies. The in cluster reconciling is run in
sync with each descheduling cycle. An in file token either changes or it
does not. If changed a new prometheus client is created. The secret
based reconciling is run async and watches for secret object changes. If a
secret changes a new client is created. The internal state of the
reconciler keeps previous connection data for clearing and checks.

The current reconciler implementation lacks mutually exclusive access.
So data races are possible. The prometheus configuration validation is
performed during every sync. The future refactorings is expected to move
the validation to the creation phase of the reconciler.

The extra unit testing is expected to cover the following scenarios:
- in cluster:
  - in file token is unchanged: no-op
  - in file token is changed: client is created or fails to be created
- secret:
  - no secret is not found: no client creation, internal state cleared
  - secret is found: if token changed a new client created, otherwise
    no-op
- prometheus config validation
- prometheus client injection

Any error during new prom client creation should be followed by closing
the previous connection and reseting the internal state. Yet, the error
handling is not that strict currently. So the current extra unit testing
keeps the incomplete testing cases as they are.

Other use of the tests is to make sure every time a new prometheus
client is created a descheduling cycle injects a new profile with the
updated prometheus clients. So the future refactoring does not introduce
a regression.
2026-02-01 18:13:18 +01:00
Jan Chaloupka
12b1ddfb52 refactor(pkg/descheduler): move prometheus client specific code under a dedicated promClientController
The underlying implementation is the same. Only moving the code under a
separate controller that can be unit test independently of the
descheduler type implementation.
2026-02-01 12:42:27 +01:00
Jan Chaloupka
fe2fb603f4 test(pkg/descheduler): test a prometheus client update propagates to a plugin profile handle 2026-01-25 22:00:28 +01:00
Jan Chaloupka
263db33052 fix(kubeClientSandbox): do not wait for pods in the fake indexers if they are already deleted 2026-01-24 14:49:38 +01:00
Jan Chaloupka
45dc5a20d3 test(kubeClientSandbox): more unit tests 2026-01-24 14:47:17 +01:00
Jan Chaloupka
f520856095 refactor(kubeClientSandbox): move the code under a separate file 2026-01-24 14:47:10 +01:00
Jan Chaloupka
e53b3d5dce refactor(pkg/descheduler): drop unsed clientset parameter from descheduler.runProfiles 2026-01-23 22:27:59 +01:00
Jan Chaloupka
770ec5affa refactor(pkg/descheduler): create fake shared informer factory only once 2026-01-23 21:34:09 +01:00
Jan Chaloupka
8f5a83279e refactor(pkg/descheduler): drop unused fakeEvictedPods variables in the unit tests 2026-01-23 15:31:57 +01:00
Jan Chaloupka
4daa7e2fbf refactor(pkg/descheduler): move prometheus setup under a helper
Prometheus is not used anywhere in the tests so there's no need to setup
it there.
2026-01-23 15:31:57 +01:00
Jan Chaloupka
433f0dbb8c refactor(pkg/descheduler): define a helper for newKubeClientSandbox with the default list of resources 2026-01-23 15:31:51 +01:00
Jan Chaloupka
ff580a0eff refactor(kubeClientSandbox): keep a cache of evicted pods and allow to reset it at the end of each descheduling cycle 2026-01-22 14:49:47 +01:00
Jan Chaloupka
4af097a806 refactor(pkg/operator): create a helper for registering indexer in the dry run mode 2026-01-22 14:44:41 +01:00
Jan Chaloupka
b3f0184af8 refactor(kubeClientSandbox): helpers for creating a node selector and node selector indexer 2026-01-22 14:44:32 +01:00
Jan Chaloupka
881ead3ed2 refactor(kubeClientSandbox): set the create pods reactor in buildSandbox 2026-01-22 14:44:14 +01:00
Jan Chaloupka
fc6d0d1132 refactor(pkg/operator): replace informerResource with a kubeClientSandbox 2026-01-22 14:41:48 +01:00
Jan Chaloupka
b6aadc1643 chore(pkg/descheduler): make TestPodEvictorReset table driven 2026-01-20 12:51:58 +01:00
Kubernetes Prow Robot
c4ec31684f Merge pull request #1802 from ingvagabund/global-node-selector-as-indexer
feat: register a node indexer for the global node selector instead of listing nodes with the selector
2026-01-12 15:08:13 +05:30
Jan Chaloupka
cf9edca33c feat(profile): inject a plugin instance ID to each built plugin 2026-01-06 12:26:35 +01:00
Jan Chaloupka
38f0f15787 chore: make gen 2026-01-04 20:23:13 +01:00
Jan Chaloupka
52f2aea444 refactor(pkg/framework/profile): add registerDefaultEvictor helper function 2026-01-04 19:43:47 +01:00
Jan Chaloupka
f3c63011cc refactor(pkg/framework/profile): add fake plugin registration helpers 2026-01-04 19:43:29 +01:00
Jan Chaloupka
47b939dd86 refactor(pkg/framework/profile): build a profile through a shared function to reduce code duplication 2026-01-04 19:42:30 +01:00