$ the-wire · showcase
Go 1.27.3 lands in kubernetes, addon-resizer ports to master
By RepoJournal · Filed · About Kubernetes · Composed from the cited sources · methodology
15 people shipped this
jm-franc
@jm-franc
1 cited
norbertcyran
@norbertcyran
1 cited
mbobrovskyi
@mbobrovskyi
1 cited
Davanum Srinivas
@dims
1 cited
nami-ant
@nami-ant
1 cited
Antoni Zawodny
@tosi3k
1 cited
huynguyennovem
@huynguyennovem
windsonsea
@windsonsea
Rub3nCT
@Rub3nCT
kubernetes-prow[bot]
@kubernetes-prow[bot]
hakman
@hakman
Yuvraj
@yindia
Mario Jason Braganza
@jasonbraganza
pohly
@pohly
serathius
@serathius
The day's most consequential change is a toolchain bump to a Go security release; everything else is cleanup, test infrastructure, and localization.
upgrade .go-version to 1.27.3 kubernetes/kubernetes
The repo's .go-version moves to go1.27.3, a security release covering crypto/x509. Committed binaries pick it up through GOTOOLCHAIN, while kube-cross and go-runner images follow once published.
Port updates from the addon-resizer-release-1.8 branch kubernetes/autoscaler
Seven years of addon-resizer development on the addon-resizer-release-1.8 branch lands on master: updated Go code, Makefile and Dockerfile build setup, README and deploy/ docs, and dependencies moved off Godeps. Anything still building addon-resizer from master was previously several years behind the branch people actually shipped.
Cancel context on termination signals kubernetes/autoscaler
registerSignalHandlers is replaced with ctrl.SetupSignalHandler, which cancels the main context and starts graceful shutdown. Previously the main context was never canceled and the process exited via os.Exit inside the signal handler; the commit also adds CapacityBuffers to the manager's schema.
Use 30/30/30 days triage intervals for Kueue PRs kubernetes/test-infra
Kueue PRs move from the org-wide 90d/30d/30d stale/rotten/close schedule to 30/30/30, driven by 329 open PRs with 98 inactive for more than 30 days. The new jobs, including ci-k8s-triage-robot-stale-prs-kueue, apply lifecycle/stale at 30d with --updated=720h.
apiserver/cacher: add slow watcher scenarios to BenchmarkSlowWatcherTax kubernetes/kubernetes
BenchmarkSlowWatcherTax gains five scenarios, including companion watchers that drain and reconnect every 500 ms, stall once, or both, at 100 events/s with the baseline and reconnect case also at 1000 events/s. No production code changes: it measures today's behavior only.
Remove group priority lookups and nil-lister forks from preemption (#142951) kubernetes/kubernetes
Preemption drops group priority lookups and nil-lister forks, and elsewhere in kubernetes the day was cleanup: deleteOnce and updateOnce extracted from etcd3, and a golangci-lint integration test that pins expected linter messages in hack/testdata/golangci-lint so a silently disabled check fails verify instead of waiting for a manual code experiment.
