Closed BenTheElder closed 1 day ago
For k8s-prow-builds (the google.com build cluster) I can't find any current mechanism auto-managing anything other than the greenhouse application image so to test turning down the instance I'm going to manually delete the service object such that the instance is soft-turned-down. If all is well we can fully turn it down later.
kubectl get -oyaml svc bazel-cache
apiVersion: v1
kind: Service
metadata:
annotations:
kubectl.kubernetes.io/last-applied-configuration: |
{"apiVersion":"v1","kind":"Service","metadata":{"annotations":{},"labels":{"run":"bazel-cache"},"name":"bazel-cache","namespace":"default"},"spec":{"ports":[{"port":8080,"protocol":"TCP"}],"selector":{"app":"greenhouse"}}}
creationTimestamp: "2018-02-06T01:25:58Z"
labels:
run: bazel-cache
name: bazel-cache
namespace: default
resourceVersion: "322348273"
selfLink: /api/v1/namespaces/default/services/bazel-cache
uid: af6fe93d-0adc-11e8-accd-42010a80009c
spec:
clusterIP: 10.63.246.72
ports:
- port: 8080
protocol: TCP
targetPort: 8080
selector:
app: greenhouse
sessionAffinity: None
type: ClusterIP
status:
loadBalancer: {}
For k8s.io the configs live here: https://github.com/kubernetes/k8s.io/tree/e18c7c6377b78a5b2935bea598e4b75497b2f89c/infra/gcp/terraform/k8s-infra-prow-build/prow-build/resources/default
cc @spiffxp
/cc
The Kubernetes project currently lacks enough contributors to adequately respond to all issues and PRs.
This bot triages issues and PRs according to the following rules:
lifecycle/stale
is appliedlifecycle/stale
was applied, lifecycle/rotten
is appliedlifecycle/rotten
was applied, the issue is closedYou can:
/remove-lifecycle stale
/lifecycle rotten
/close
Please send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle stale
The Kubernetes project currently lacks enough active contributors to adequately respond to all issues and PRs.
This bot triages issues and PRs according to the following rules:
lifecycle/stale
is appliedlifecycle/stale
was applied, lifecycle/rotten
is appliedlifecycle/rotten
was applied, the issue is closedYou can:
/remove-lifecycle rotten
/close
Please send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle rotten
/remove-lifecyle rotten /assign @ameukam
I still see some cache.
/data/kubernetes/kubernetes,b7fa00de3a7c5dec1bb841e81f144ae9 # du -d 1 -h
2.6T ./cas
555.1M ./ac
2.6T .
It's possible it's stale cache. I'll manually clear the cache and see what's happening.
Greenhouse is an LRU cache so it will always be ~full unless it has zero usage since setup.
With a full cache we could check how recently entries were accessed.
Greenhouse is an LRU cache so it will always be ~full unless it has zero usage since setup.
With a full cache we could check how recently entries were accessed.
Good idea!
/data/kubernetes/kubernetes,b7fa00de3a7c5dec1bb841e81f144ae9/cas # ls -lhur | tail -20
-rw------- 1 root root 129 Mar 3 15:08 af72d6d45af49bc12c8f24988cda5bbcab607eb1acd0ee0a16017de8329de1c9
-rw------- 1 root root 163.6M Mar 3 15:08 91f1cd8a5ec9355c6131100df2613216874926fe016331cc275e63e9d8badc47
-rw------- 1 root root 41 Mar 3 15:08 6f9f6f97af118c28ca02aee095cfb9f5c0642990b8bc0a5a66207043f80436bb
-rw------- 1 root root 154.5M Mar 3 15:08 54296156a8a8f5489f82dc67e220bd1414e81aeb24c51b2fbb557d0490568a73
-rw------- 1 root root 33 Mar 3 15:08 2d5f8cdd41a08fb0227bef2feb8afdbf5c3221d2162740deddac689099411c6a
-rw------- 1 root root 953.1M Mar 3 15:08 ff8c981b8dea03e4619cee094b5a515c949f166d5583c28dfc450330aaa242d5
-rw------- 1 root root 33 Mar 3 15:08 ad6e165a5b018bb213d3e40d5059797922ffc119398e6eb0c8412d55dc9c0da9
-rw------- 1 root root 129 Mar 3 15:08 a5b42259f179a25a0588cf0f157e8b69f911d746d0fbdc1e2b0419a0c5c67533
-rw------- 1 root root 41 Mar 3 15:08 70238e0aaa38087e5f3ff632e6c3a761acd50227a0a745296d1e3282f0975860
-rw------- 1 root root 318.9M Mar 3 15:08 59f24f968b5d6e141ff700f97ec99a4958e89a25d3ac96a66fd2d16f8138eab2
-rw------- 1 root root 41 Mar 3 15:08 c2befdb860da789db1e8c73c0e209a70d3c9e3861e9e037dd11cc5c9dfe7c782
-rw------- 1 root root 33 Mar 3 15:08 15c45ff20039d7024b9a7bec6f6fa0cf14981fc1643c2ff0e995f7b8271ac4a7
-rw------- 1 root root 129 Mar 3 15:08 0b22393597be5fed56bdf53584bb177cc7eb91f9ca397ce47727c5d7b287b4b3
-rw------- 1 root root 1.1G Mar 3 15:08 827675094d35b9b5d06bda7e8e75c1aa3a5f83b84f4b12a5bfdf149d1cba9606
-rw------- 1 root root 453.9M Mar 3 15:08 b09b0db9eec6067069704f8606ee16488f142c2fa7cb8ceba7b0edd8054dad56
-rw------- 1 root root 0 Mar 3 15:09 e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
-rw------- 1 root root 33 Mar 3 15:09 e2dc208d675b1645afef7fcae438debb5d12cefb5152e01efa0c118eb1d201e2
-rw------- 1 root root 129 Mar 3 15:09 be629e40e1a2fe0c8a22d060cb0177f53523a6958c7c366935a52658b4b2a923
-rw------- 1 root root 41 Mar 3 15:09 483ab076e9a4a36a4184a1c8a6ec82e0bc9453b9a17e6b4dc96af9589320c9ba
-rw------- 1 root root 60.1K Mar 23 20:58 7fffc697b4402e1f9c9cbdcc8c0cfae028c0fc2fdf7cc63eedbaaed57b075e6e
We can use GCS as a bazel cache https://docs.bazel.build/versions/main/remote-caching.html#google-cloud-storage
We can use GCS as a bazel cache https://docs.bazel.build/versions/main/remote-caching.html#google-cloud-storage
We don't want to do that because garbage collection is ?? Greenhouse basically exists so we can do LRU of bounded size.
The Kubernetes project currently lacks enough contributors to adequately respond to all issues and PRs.
This bot triages issues and PRs according to the following rules:
lifecycle/stale
is appliedlifecycle/stale
was applied, lifecycle/rotten
is appliedlifecycle/rotten
was applied, the issue is closedYou can:
/remove-lifecycle stale
/lifecycle rotten
/close
Please send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle stale
I think we're ready to do this? /remove-lifecycle stale
The Kubernetes project currently lacks enough contributors to adequately respond to all issues and PRs.
This bot triages issues and PRs according to the following rules:
lifecycle/stale
is appliedlifecycle/stale
was applied, lifecycle/rotten
is appliedlifecycle/rotten
was applied, the issue is closedYou can:
/remove-lifecycle stale
/lifecycle rotten
/close
Please send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle stale
The Kubernetes project currently lacks enough active contributors to adequately respond to all issues and PRs.
This bot triages issues and PRs according to the following rules:
lifecycle/stale
is appliedlifecycle/stale
was applied, lifecycle/rotten
is appliedlifecycle/rotten
was applied, the issue is closedYou can:
/remove-lifecycle rotten
/close
Please send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle rotten
/remove-lifecycle rotten /lifecycle frozen
Completely forgot about this. The infrastructure and code are gone.
/close
@ameukam: Closing this issue.
Part of https://github.com/kubernetes/enhancements/issues/2420
test-infra has used RBE instead for some time now, greenhouse was used for Kubernetes builds but no supported branches are anymore. Greenhouse is a pretty large deployment for test-infra (dedicated 32 core VM, 3TB pd-ssd per build cluster), which is rather unjustified without Kubernetes using it anymore.
Additionally, it doesn't appear to be properly auto-deployed anymore? and isn't actively developed / needed by the project at large. We should just turn it down, its time has passed.
Any remaining projects that happen to be using it should fallback to bazel without caching automatically and if they find that they need a cache still they can either spin up a more reasonably sized deployment with SIG k8s-infra (greenhouse is well documented and we'll leave the sources for now) or use some alternate mechanism (e.g. RBE).
/assign