• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

kubernetes-sigs / lws / 42349654702
88%

Build:
DEFAULT BRANCH: main
Ran 07 Nov 2026 07:08AM UTC
Jobs 1
Files 41
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

07 Nov 2026 07:03AM UTC coverage: 88.213%. Remained the same
42349654702

push

github

web-flow
Add a scalability test harness for LeaderWorkerSet (#1216)

* Add a scalability test harness for LeaderWorkerSet

Add test/performance with a runner that generates LeaderWorkerSets from
a config, creates them and runs the configured scaling phases, and
writes summary.yaml with the wall time, throughput and latency
percentiles of each phase. A LeaderWorkerSet is done when its status
for the current generation shows exactly the target replicas, all
ready and Available, read from an informer cache. After each phase the
runner checks that every LeaderWorkerSet has exactly replicas * size
pods, all Running and Ready, and no others. With a range spec the run
fails when a measurement is above its bound.

make test-performance runs it through hack/e2e-test.sh with
PERFORMANCE_TEST=true: the e2e kind cluster, with its own kubeconfig,
plus simulated KWOK nodes, so pods become Ready without kubelets while
the real API server, scheduler and LWS controller carry the load. The
baseline workload has many small and a few large LeaderWorkerSets in
both group identity modes. The objects are built with test/wrappers.

Rolling updates, group recreation, deletion and the periodic CI job
are left for follow-ups.

Refs #1146

* Report controller CPU, memory and API requests in the scalability test

The runner now reads the metrics endpoint of each LWS controller pod
through kubectl port-forward, with a token of the controller service
account, and adds to each phase of summary.yaml the CPU seconds, the
peak resident memory and the API requests by method of the controller
pods during that phase. The pods are summed, since the leader
reconciles and every pod serves webhooks. Before the first phase the
runner also waits until a pod holds the leader lease, so the election
does not count as latency.

The range spec can bound these as controller.cpuSeconds,
controller.peakMemoryMiB and controller.apiRequests.

Refs #1146

* Address review on the scalability test runner

* Keep the... (continued)

10784 of 12225 relevant lines covered (88.21%)

31749.91 hits per line

Jobs
ID Job ID Ran Files Coverage
1 42349654702.1 07 Nov 2026 07:08AM UTC 41
88.21
GitHub Action Run
Source Files on build 42349654702
  • Tree
  • List 41
  • Changed 0
  • Source Changed 0
  • Coverage Changed 0
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • 97c376c8 on github
  • Prev Build on main (#42237990636)
  • Next Build on main (#42400723935)
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc