Commit Graph
685 Commits
Author SHA1 Message Date
Simone Tiraboschi 3a608a590a feat(eviction): add annotations to eviction requests for observability
Although eviction requests (policy/v1) are not persisted long term,
their API still implements the full metav1.ObjectMeta struct. While
name and namespace refer to the pod being evicted, eviction requests
can still carry annotations.

This change adds annotations to descheduler-initiated evictions,
including the requester, reason, and the strategy or plugin that
triggered them.

While these details are already logged by the descheduler, exposing them
as annotations allows external webhooks or controllers to provide
clearer context about each eviction request, both for tracking and
prioritization purposes.

Signed-off-by: Simone Tiraboschi <stirabos@redhat.com>
2025-10-13 12:49:03 +02:00
S J Tharun 7d6f6fedec fix gofmt error 2025-10-13 08:29:27 +05:30
S J Tharun 3033aec6a0 descheduler: handle single-node clusters gracefully 2025-10-11 07:53:41 +05:30
Emin Aktas d47e077897 fix: correct comment for HaveNoEvictionAnnotation 2025-09-20 12:53:56 +03:00
googs1025 257bd55909 standardize protectionEnabled param 2025-08-19 17:52:54 +08:00
Kubernetes Prow RobotandGitHub 63ecaec0ef Merge pull request #1729 from googs1025/validate/improve
Improves the validation logic in the plugins
2025-08-15 03:31:07 -07:00
Kubernetes Prow RobotandGitHub e5c57a759b Merge pull request #1603 from googs1025/feature/add_dra_arg
feature: add PodsWithResourceClaims parameter in DefaultEvictorArgs PodProtections
2025-08-13 06:57:10 -07:00
googs1025 9c7e01de67 feature: add PodsWithResourceClaims parameter in DefaultEvictorArgs PodProtections
Signed-off-by: googs1025 <googs1025@gmail.com>
2025-08-13 20:35:36 +08:00
googs1025 e5ea03ce75 add ValidateHighNodeUtilizationPluginConfig unit test
Signed-off-by: googs1025 <googs1025@gmail.com>
2025-08-13 17:15:21 +08:00
googs1025 2cce141feb Improves the validation logic in the plugins
Signed-off-by: googs1025 <googs1025@gmail.com>
2025-08-07 17:21:01 +08:00
googs1025 d9697591d5 add EvictionProtection filed for DefaultEvictorArgs
Signed-off-by: googs1025 <googs1025@gmail.com>
2025-08-04 19:34:16 +08:00
Jan Chaloupka 7380aa6e0a DefaultEvictor: introduce no-eviction policy
NoEvictionPolicy dictates whether a no-eviction policy is prefered or mandatory.
Needs to be used with caution as this will give users ability to protect their pods
from eviction. Which might work against enfored policies. E.g. plugins evicting pods
violating security policies.
2025-07-24 15:47:05 +02:00
Jan Chaloupka d0548b75d7 TestSortPodsBasedOnPriorityLowToHigh: check the whole sorted list of pods 2025-07-24 12:48:03 +02:00
Jan Chaloupka 6e9d8891c5 defaultevictor_test.go: replace descheduler.alpha.kubernetes.io/evict literal with evictPodAnnotationKey const 2025-07-24 12:48:00 +02:00
Jan Chaloupka 5cc9e68127 Drop assignment if default test values 2025-07-24 12:04:07 +02:00
Amir Alavi eadfe4a546 fix: topologyspreadconstraint plugin to not add PodNodeAffinity unless the inclusion policy is honor 2025-07-10 17:42:57 -04:00
Amir Alavi fbf11df729 fix: topologyspreadconstraint to prefer evictable before sorting domains
Sort pods that are above ideal avg based on the criteria that they fit on other nodes that are below avg

Signed-off-by: Amir Alavi <amiralavi7@gmail.com>
2025-07-10 17:42:12 -04:00
Ricardo Maraschini 6e714a2134 chore: stop with no-op default evictor settings
we have been carrying these no-op for quite a while now. we should only
set defaults when they are different from what they are being provided
by the user.
2025-07-07 15:31:49 +02:00
Jan Chaloupka 89eab59d82 logger: Align with the previous logger verbosity 2025-06-25 09:27:57 +02:00
Kubernetes Prow RobotandGitHub ace001c618 Merge pull request #1655 from googs1025/feature/contextual_logging
feature: use contextal logging for plugins
2025-06-25 00:16:29 -07:00
googs1025 33894afe2b feature: use contextal logging for plugins
Signed-off-by: googs1025 <googs1025@gmail.com>
2025-06-13 19:45:55 +08:00
Ricardo Maraschini a2b899aa15 feat(prometheus): allow different url schemes
as per prometheus golang client implementation: the only url validation
done is by means of an `url.Parse()` call. we should do the same and not
enforce the usage of https scheme.

our readme even shows an example of descheduler config using http
prometheus url scheme.
2025-06-12 11:23:09 +02:00
lowang-bh b63b09089e metrics name refact and mark old one deprecated
Signed-off-by: lowang-bh <lhui_wang@163.com>
2025-06-02 14:13:46 +08:00
Martin Tapp e06443ef40 *1677 Allow Succeeded and Failed states in PodLifeTime 2025-05-23 15:56:37 -04:00
Kubernetes Prow RobotandGitHub 9f918371a2 Merge pull request #1694 from ingvagabund/plugins-new-context
Extend plugin's New with a context.Context
2025-05-19 05:31:14 -07:00
Jan Chaloupka 1974c12e0f Extend plugin's New with a context.Context
The new context.Context can be later used for passing a contextualized
logger. Or, other initialization steps that require the context.
2025-05-19 12:23:44 +02:00
googs1025 b3aeca73db refactor: separate eviction constraints to constraints.go
Signed-off-by: googs1025 <googs1025@gmail.com>
2025-05-19 18:02:41 +08:00
Kubernetes Prow RobotandGitHub d34848086c Merge pull request #1691 from googs1025/fix/listall
fix(example): list only active pod
2025-05-12 10:55:14 -07:00
Kubernetes Prow RobotandGitHub 9aa6d79c21 Merge pull request #1688 from ingvagabund/plugins-taints-do-not-list-all-pods
RemovePodsViolatingNodeTaints: list only pods that are not failed/suceeded
2025-05-12 06:55:18 -07:00
googs1025 7a76d9f0d3 fix(RemovePodsViolatingNodeTaints): list only active pod
Signed-off-by: googs1025 <googs1025@gmail.com>
2025-05-12 21:45:57 +08:00
Kubernetes Prow RobotandGitHub 71746262b1 Merge pull request #1684 from googs1025/refactor_topology
chore: move namespaces filtering logic to New()
2025-05-11 12:29:14 -07:00
Jan Chaloupka 6691720da5 RemovePodsViolatingNodeTaints: list only pods that are not failed/suceeded
Listing pods was incorrectly changed to listing all pods during code
refactoring.
2025-05-10 21:12:06 +02:00
googs1025 0a691debfb feature: sort pods by restarts count in RemovePodsHavingTooManyRestarts plugin 2025-05-09 13:15:38 +08:00
googs1025 fbc875fac1 chore: move namespaces filtering logic to New() 2025-05-07 19:47:30 +08:00
googs1025 957c5bc8e0 optimize: NodeFit function by reordering checks for performance 2025-05-05 21:09:14 +08:00
Ricardo Maraschini 35a7178df6 feat: introduce strict eviction policy
with strict eviction policy the descheduler only evict pods if the pod
contains a request for the given threshold. for example, if using a
threshold for an extended resource called `example.com/gpu` only pods
who request such a resource will be evicted.
2025-04-09 16:48:32 +02:00
Kubernetes Prow RobotandGitHub 17b90969cf Merge pull request #1629 from googs1025/fix_sortDomains
fix removepodsviolatingtopologyspreadconstraint plugin sort logic
2025-03-28 09:06:44 -07:00
googs1025 b4b203cc60 chore: add unit test for sortDomains func 2025-03-28 20:58:54 +08:00
googs1025 59d1d5d1b9 making isEvictable and hasSelectorOrAffinity invoked only once
Signed-off-by: googs1025 <googs1025@gmail.com>
2025-03-28 20:57:02 +08:00
Ricardo Maraschini 98e6ed6587 chore: log average and assessed thresholds
when calculating the average an applying the deviations it would be nice
to also see the assessed values.

this commit makes the descheduler logs these values when using level 3.
2025-03-28 10:16:20 +01:00
googs1025 4548723dea fix(removepodsviolatingtopologyspreadconstraint): fix removepodsviolatingtopologyspreadconstraint plugin sort logic
Signed-off-by: googs1025 <googs1025@gmail.com>
2025-03-27 23:49:45 +08:00
Kubernetes Prow RobotandGitHub 9b9ae9a3be Merge pull request #1650 from ingvagabund/nodeutilization-deviation-skip-nodes-without-extended-resource
[nodeutilization]: skip nodes without extended resource when computing the average utilization
2025-03-20 05:58:31 -07:00
Jan Chaloupka c22d773200 [nodeutilization] test nodes without extended resource when computing the average utilization 2025-03-20 13:22:28 +01:00
Ricardo Maraschini 95a631f6a5 feat: move classifier to its own package
move the classifier to its own package. introduces a generic way of
classifying usages against thresholds.
2025-03-20 11:02:47 +01:00
Ricardo Maraschini 54d0a22ad1 chore: comment the code and simplify some things
this commit comments the low and high utilization plugins. this also
simplifies a little bit where it was possible without affecting too
much.
2025-03-20 10:06:45 +01:00
Ricardo Maraschini 87ba84b2ad feat: refactoring thresholds and usage assessment
this commit refactors the thresholds and usage assessment for the node
utilization plugins. both high and low plugins are affected by this
change.
2025-03-19 17:38:41 +01:00
Jan Chaloupka 04ebdbee32 [nodeutilization]: produce node utilization of resources that are listed in the list of resources 2025-03-19 12:35:19 +01:00
Jan Chaloupka e283c31030 [nodeutilization]: prometheus usage client with prometheus metrics 2025-03-17 16:25:17 +01:00
Kubernetes Prow RobotandGitHub be4abe1727 Merge pull request #1614 from ingvagabund/nodeutilization-metrics-source
[nodeutilization]: allow to set a metrics source as a string so it can be later extended for exclusive configuration
2025-03-14 01:57:47 -07:00
Jan Chaloupka e14b86eb8c [nodeutilization]: allow to set a metrics source as a string so it can be later extended for exclusive configuration 2025-03-13 18:00:27 +01:00