Files
actions-runner-controller/acceptance/testdata/runnerset.envsubst.yaml
Yusuke Kuoka b5194fd75a Enhance RunnerSet to optionally retain PVs accross restarts (#1340)
* Enhance RunnerSet to optionally retain PVs accross restarts

This is our initial attempt to bring back the ability to retain PVs across runner pod restarts when using RunnerSet.
The implementation is composed of two new controllers, `runnerpersistentvolumeclaim-controller` and `runnerpersistentvolume-controller`.
It all starts from our existing `runnerset-controller`. The controller now tries to mark any PVCs created by StatefulSets created for the RunnerSet.
Once the controller terminated statefulsets, their corresponding PVCs are clean up by `runnerpersistentvolumeclaim-controller`, then PVs are unbound from their corresponding PVCs by `runnerpersistentvolume-controller` so that they can be reused by future PVCs createf for future StatefulSets that shares the same same StorageClass.

Ref #1286

* Update E2E test suite to cover runner, docker, and go caching with RunnerSet + PVs

Ref #1286
2022-05-16 09:26:48 +09:00

226 lines
7.1 KiB
YAML

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: ${NAME}
# In kind environments, the provider writes:
# /var/lib/docker/volumes/KIND_NODE_CONTAINER_VOL_ID/_data/local-path-provisioner/PV_NAME
# It can be hundreds of gigabytes depending on what you cache in the test workflow. Beware to not encounter `no space left on device` errors!
# If you did encounter no space errorrs try:
# docker system prune
# docker buildx prune #=> frees up /var/lib/docker/volumes/buildx_buildkit_container-builder0_state
# sudo rm -rf /var/lib/docker/volumes/KIND_NODE_CONTAINER_VOL_ID/_data/local-path-provisioner #=> frees up local-path-provisioner's data
provisioner: rancher.io/local-path
reclaimPolicy: Retain
volumeBindingMode: WaitForFirstConsumer
---
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: ${NAME}-var-lib-docker
labels:
content: ${NAME}-var-lib-docker
provisioner: rancher.io/local-path
reclaimPolicy: Retain
volumeBindingMode: WaitForFirstConsumer
---
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: ${NAME}-cache
labels:
content: ${NAME}-cache
provisioner: rancher.io/local-path
reclaimPolicy: Retain
volumeBindingMode: WaitForFirstConsumer
---
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: ${NAME}-runner-tool-cache
labels:
content: ${NAME}-runner-tool-cache
provisioner: rancher.io/local-path
reclaimPolicy: Retain
volumeBindingMode: WaitForFirstConsumer
---
apiVersion: actions.summerwind.dev/v1alpha1
kind: RunnerSet
metadata:
name: ${NAME}
spec:
# MANDATORY because it is based on StatefulSet: Results in a below error when omitted:
# missing required field "selector" in dev.summerwind.actions.v1alpha1.RunnerSet.spec
selector:
matchLabels:
app: ${NAME}
# MANDATORY because it is based on StatefulSet: Results in a below error when omitted:
# missing required field "serviceName" in dev.summerwind.actions.v1alpha1.RunnerSet.spec]
serviceName: ${NAME}
#replicas: 1
# From my limited testing, `ephemeral: true` is more reliable.
# Seomtimes, updating already deployed runners from `ephemeral: false` to `ephemeral: true` seems to
# result in queued jobs hanging forever.
ephemeral: ${TEST_EPHEMERAL}
enterprise: ${TEST_ENTERPRISE}
group: ${TEST_GROUP}
organization: ${TEST_ORG}
repository: ${TEST_REPO}
#
# Custom runner image
#
image: ${RUNNER_NAME}:${RUNNER_TAG}
#
# dockerd within runner container
#
## Replace `mumoshu/actions-runner-dind:dev` with your dind image
#dockerdWithinRunnerContainer: true
dockerdWithinRunnerContainer: ${RUNNER_DOCKERD_WITHIN_RUNNER_CONTAINER}
#
# Set the MTU used by dockerd-managed network interfaces (including docker-build-ubuntu)
#
#dockerMTU: 1450
#Runner group
# labels:
# - "mylabel 1"
# - "mylabel 2"
labels:
- "${RUNNER_LABEL}"
#
# Non-standard working directory
#
# workDir: "/"
template:
metadata:
labels:
app: ${NAME}
spec:
containers:
- name: runner
imagePullPolicy: IfNotPresent
env:
- name: RUNNER_FEATURE_FLAG_EPHEMERAL
value: "${RUNNER_FEATURE_FLAG_EPHEMERAL}"
- name: GOMODCACHE
value: "/home/runner/.cache/go-mod"
volumeMounts:
# Cache docker image layers, in case dockerdWithinRunnerContainer=true
- name: var-lib-docker
mountPath: /var/lib/docker
# Cache go modules and builds
# - name: gocache
# # Run `goenv | grep GOCACHE` to verify the path is correct for your env
# mountPath: /home/runner/.cache/go-build
# - name: gomodcache
# # Run `goenv | grep GOMODCACHE` to verify the path is correct for your env
# # mountPath: /home/runner/go/pkg/mod
- name: cache
# go: could not create module cache: stat /home/runner/.cache/go-mod: permission denied
mountPath: "/home/runner/.cache"
- name: runner-tool-cache
# This corresponds to our runner image's default setting of RUNNER_TOOL_CACHE=/opt/hostedtoolcache.
#
# In case you customize the envvar in both runner and docker containers of the runner pod spec,
# You'd need to change this mountPath accordingly.
#
# The tool cache directory is defined in actions/toolkit's tool-cache module:
# https://github.com/actions/toolkit/blob/2f164000dcd42fb08287824a3bc3030dbed33687/packages/tool-cache/src/tool-cache.ts#L621-L638
#
# Many setup-* actions like setup-go utilizes the tool-cache module to download and cache installed binaries:
# https://github.com/actions/setup-go/blob/56a61c9834b4a4950dbbf4740af0b8a98c73b768/src/installer.ts#L144
mountPath: "/opt/hostedtoolcache"
# Valid only when dockerdWithinRunnerContainer=false
- name: docker
volumeMounts:
# Cache docker image layers, in case dockerdWithinRunnerContainer=false
- name: var-lib-docker
mountPath: /var/lib/docker
# image: mumoshu/actions-runner-dind:dev
# For buildx cache
- name: cache
mountPath: "/home/runner/.cache"
volumeClaimTemplates:
- metadata:
name: vol1
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 10Mi
storageClassName: ${NAME}
## Dunno which provider supports auto-provisioning with selector.
## At least the rancher local path provider stopped with:
## waiting for a volume to be created, either by external provisioner "rancher.io/local-path" or manually created by system administrator
# selector:
# matchLabels:
# runnerset-volume-id: ${NAME}-vol1
- metadata:
name: vol2
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 10Mi
storageClassName: ${NAME}
# selector:
# matchLabels:
# runnerset-volume-id: ${NAME}-vol2
- metadata:
name: var-lib-docker
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 10Mi
storageClassName: ${NAME}-var-lib-docker
- metadata:
name: cache
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 10Mi
storageClassName: ${NAME}-cache
- metadata:
name: runner-tool-cache
# It turns out labels doesn't distinguish PVs across PVCs and the
# end result is PVs are reused by wrong PVCs.
# The correct way seems to be to differentiate storage class per pvc template.
# labels:
# id: runner-tool-cache
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 10Mi
storageClassName: ${NAME}-runner-tool-cache
---
apiVersion: actions.summerwind.dev/v1alpha1
kind: HorizontalRunnerAutoscaler
metadata:
name: ${NAME}
spec:
scaleTargetRef:
kind: RunnerSet
name: ${NAME}
scaleUpTriggers:
- githubEvent:
workflowJob: {}
amount: 1
duration: "10m"
minReplicas: ${RUNNER_MIN_REPLICAS}
maxReplicas: 10
scaleDownDelaySecondsAfterScaleOut: ${RUNNER_SCALE_DOWN_DELAY_SECONDS_AFTER_SCALE_OUT}