Open zickzackv opened 9 months ago
zickzackv
commented
8 months ago We now setup our arc (Kubernetes Runners) with a DIND setup. So the runner container has a docker binary and a docker socket available and one can start other containers with that docker daemon and socket. Sadly the socket's path is specified within an environment variable; that environment is lost when executing sudo (without -E). Root (with sudo) will not find the docker socket to start an nix daemon.
zickzackv
commented
8 months ago Even if the nix daemon could be started within our ARC/DIND setup I would suspect that a nix deamon running in another pod/container will not have access to /nix/store in the original container/pod where the installer run.
remi-gelinas
commented
8 months ago Yep, same ARC Kubernetes setup for us and same issue and logs/error code. Until this is solved we'll have to go back to Cachix's install action
lucperkins
commented
6 months ago @zickzackv Could you give it another try with a more recent commit or on plain old main? We've overhauled this Action pretty significantly in recent weeks.
grahamc
commented
6 months ago As an aside, we'd really like to support this better -- but we don't have anyone that we're regularly interacting with to do design & implementation reviews with. If someone is able to do that, I'd be glad to be in touch: gc@determinate.systems.
grahamc
commented
6 months ago @remi-gelinas / @zickzackv -- if possible, can you paste sanitized env dumps? I'm curious what environment variables are present that we might use to turn on and off relevant behavior.
zickzackv
commented
5 months ago @grahamc Sadly I can't tests the installer under ARC. I'm not working anymore with ARC :-(
remi-gelinas
commented
5 months ago @grahamc I think I can grab a sanitized dump for you. No clue if it'll be helpful, but I'll see what I can extract today
remi-gelinas
commented
5 months ago @grahamc Here is a heavily sanitized env dump from one of our runners attempting to use v11 of the action:
KUBERNETES_SERVICE_PORT_HTTPS=443
GITHUB_WORKSPACE=/runner/_work/runner/runner
KUBERNETES_SERVICE_PORT=443
HOSTNAME=<redacted>
RUNNER_WORKDIR=/runner/_work
GITHUB_ACTION=__run
DOCKERD_IN_RUNNER=false
GITHUB_RUN_NUMBER=186
RUNNER_NAME=<redacted>
GITHUB_REPOSITORY_OWNER_ID=<redacted>
GITHUB_URL=https://github.com/
RUNNER_ORG=<redacted>
AWS_DEFAULT_REGION=us-west-2
ACTIONS_RUNNER_HOOK_JOB_COMPLETED=/etc/arc/hooks/job-completed.sh
GITHUB_TRIGGERING_ACTOR=<actor>
GITHUB_REF_TYPE=branch
AWS_REGION=us-west-2
DOCKER_ENABLED=true
GITHUB_ACTIONS=true
_=/usr/bin/env
RUNNER_ENVIRONMENT=self-hosted
GITHUB_REF=refs/pull/72/merge
RUNNER_OS=Linux
GITHUB_REF_PROTECTED=false
RUNNER_STATUS_UPDATE_HOOK=false
HOME=/home/runner
GITHUB_API_URL=https://api.github.com
RUNNER_TRACKING_ID=<trackingid>
RUNNER_ARCH=X64
RUNNER_TEMP=/runner/_work/_temp
GITHUB_EVENT_PATH=/runner/_work/_temp/_github_workflow/event.json
GITHUB_EVENT_NAME=pull_request
GITHUB_ACTIONS_RUNNER_EXTRA_USER_AGENT=actions-runner-controller/v0.27.5
RUNNER_ASSETS_DIR=/runnertmp
GITHUB_RUN_ID=<runid>
DISABLE_RUNNER_UPDATE=true
GITHUB_ACTOR=<actorname>
RUNNER_EPHEMERAL=true
GITHUB_RUN_ATTEMPT=1
GITHUB_GRAPHQL_URL=https://api.github.com/graphql
RUNNER_GROUP=
GITHUB_SERVER_URL=https://github.com
SHLVL=4
KUBERNETES_PORT_443_TCP_PROTO=tcp
GITHUB_ACTOR_ID=101587823
RUNNER_TOOL_CACHE=/opt/hostedtoolcache
GITHUB_JOB=<jobname>
AWS_STS_REGIONAL_ENDPOINTS=regional
ACTIONS_RUNNER_HOOK_JOB_STARTED=/etc/arc/hooks/job-started.sh
DOCKER_HOST=unix:///run/docker.sock
GITHUB_REPOSITORY=<reponame>
GITHUB_RETENTION_DAYS=30
RUNNER_WORKSPACE=/runner/_work/runner
GITHUB_ACTION_REPOSITORY=
PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/home/runner/.local/bin/
GITHUB_BASE_REF=main
CI=true
ImageOS=ubuntu20
GITHUB_REPOSITORY_OWNER=<redacted>
GITHUB_HEAD_REF=<branchname>
GITHUB_ACTION_REF=
RUNNER_LABELS=self-hosted,np,default
RUNNER_ENTERPRISE=
GITHUB_WORKFLOW=<workflowname>
DEBIAN_FRONTEND=noninteractive
OLDPWD=/Maybe of note are the DOCKER_ENABLED and DOCKERD_IN_RUNNER vars, but I have no idea. Let me know if I can help further, or if you would like a guinea pig to test a fix.
I wanted to try out the installer action on our recently setup actions-runner-controller. Sadly it failed while setting up the docker shim.
docker-runner: 2.312.0 actions-runner-controller: 0.27.5
I could't trace the line in src/main.ts where docker is failing.