koord-descheduler: LowNodeLoad supports anomaly detector

[eahydra]

Signed-off-by: Joseph joseph.t.lee@outlook.com

Ⅰ. Describe what this PR does

One of the commonly encountered scenarios: the load of the node suddenly rises, but then drops back quickly. If LowNodeLoad happens to be executing at this time, an unexpected eviction action is performed.

Therefore, in order to avoid such sudden scenarios, LowNodeLoad introduces an anomaly detector to help determine whether a node is really overutilized. The default detection is that if the load of a node is higher than HighThresholds for five consecutive times, then the node is considered really overutilized.

Users can configure the detection thresholds via LowNodeArgs.AnomalyThresholds.

Ⅱ. Does this pull request fix one issue?

Ⅲ. Describe how to verify it

1. create stress pods
2. run the latest descheduler with LowNodeLoad plugin.
3. watch the logs, you can see the logs:

I0206 00:52:33.366587   14886 low_node_load.go:184] "None of the nodes were detected as anomalous, nothing to do here"

4. After run LowNodeLoad.Balance five times, the stress pods on the nodes that are overutilized will be evicted.

Ⅳ. Special notes for reviews

V. Checklist

• I have written necessary docs and comments
• I have added necessary unit tests and integration tests
• All checks passed in make test

[codecov]

Codecov Report

Base: 66.99% // Head: 67.60% // Increases project coverage by +0.60% 🎉

Coverage data is based on head (cd901fa) compared to base (1e3fa53).
Patch coverage: 83.14% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #987      +/-   ##
==========================================
+ Coverage   66.99%   67.60%   +0.60%     
==========================================
  Files         244      244              
  Lines       28330    28140     -190     
==========================================
+ Hits        18981    19025      +44     
+ Misses       8049     7816     -233     
+ Partials     1300     1299       -1     

Flag	Coverage Δ
unittests	67.60% <83.14%> (+0.60%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...ler/apis/config/validation/validation_loadaware.go	0.00% <0.00%> (ø)
pkg/descheduler/utils/anomaly/types.go	0.00% <0.00%> (ø)
...duler/framework/plugins/loadaware/low_node_load.go	73.44% <85.96%> (+2.85%)	⬆️
pkg/descheduler/utils/anomaly/basic_detector.go	89.24% <89.24%> (ø)
pkg/descheduler/utils/anomaly/counter.go	100.00% <100.00%> (ø)
pkg/koordlet/executor/resource_update_executor.go
pkg/koordlet/executor/types.go
pkg/koordlet/executor/resctrl_executor.go
pkg/koordlet/runtimehooks/reconciler/reconciler.go	67.05% <0.00%> (+1.17%)	⬆️
pkg/scheduler/plugins/coscheduling/core/gang.go	78.23% <0.00%> (+2.15%)	⬆️
... and 1 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[saintube]

/lgtm

[zwzhang0107]

/lgtm

[jasonliu747]

/lgtm

[FillZpp]

/lgtm

[koordinator-bot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: FillZpp, jasonliu747

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• pkg/descheduler/OWNERS [FillZpp]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment