Skip to content

Latest commit

 

History

History
66 lines (37 loc) · 4.62 KB

README.md

File metadata and controls

66 lines (37 loc) · 4.62 KB
Raw

devstats-k8s-lf

DevStats kubernetes deployment for LF projects (bare kubernetes, no Helm packaging) - this also includes kubernetes util scripts to manipulate DevStats deployment

Set default namespace

You should set namespace to 'devstats' first: ./switch_namespace.sh devstats.

Images

See github.com/cncf/devstats-docker-images for the info about how to build docker images used by this deployment. You can also test them there.

Debugging running pods

You can run shell inside the running container:

  • To bash into a running pod do: ./util/pod_shell.sh pod-name, where pod name can be for example: devstats-provision-cncf-1550826466080940119.

Secrets

Secret data is not checked-in into the repository. For each file in ./kubernetes/secrets/*.secret.example you need to create your own ./kubernetes/secrets/*.secret and propagate into your cluser.

Once all those files are created, use ./kubernetes/create_secrets.sh script to propagate them into you Kubernetes cluster.

Please note that vi automatically adds new line to all text files, to remove it run truncate -s -1 on a saved file.

Test pods before actually running them

This is optional (it starts real containers with their command replaced with the shell and then you can use ./kubernetes/pod_shell.sh to shell into them and run their actual command manually):

  • To dry-run manifests against your Kubernetes cluster, run: ./kubernetes/dryrun_manifest.sh ./kubernetes/manifests/*.yaml. This will show the final manifest and check if it can be applied.
  • Prepend ./kubernetes/apply_manifest.sh call with SLEEP=1 to apply a manifest with its command replaced with sleep 36000. You can the use ./util/pod_shell.sh to bash into that container running forever sleep and check its state. You should delete that pod after testing it. It will terminate automatically after 10 hours (36000s).
  • Observe environment via: env | grep -E '(GHA2DB|^PG_|^ES_|^ONLY|INIT|^PROJ)' | sort. See PV mounts: df -h. Delete pod kubectl delete pod pod-name.

Deploy on Kubernetes

  • Run ./kubernetes/apply_manifest.sh ./kubernetes/manifests/devstats-pvc.yaml to create presistent volume and persisten volume claim for git repository clones storage. This must be done first.
  • You need all secrets to be populated via ./kubernetes/create-secrets.sh. There is a ./kubernetes/manifests/devstats-secrets.yaml file, but all secret values are eHh4Cg== here (this is xxx base 64 encoded).
  • Run PROJ=... PROJDB=... PROJREPO=... INIT=1 ./kubernetes/apply_manifest.sh ./kubernetes/manifests/devstats-provision.yaml to do an initial Kubernetes deployment (bootstraps logs database, users and deploys first project). You can use ONLYINIT=1 to test bootstraping logs database and users only - that will skip actual first project provision.
  • Run PROJ=... PROJDB=... PROJREPO=... ./kubernetes/apply_manifest.sh ./kubernetes/manifests/devstats-provision.yaml to deploy any next project.
  • Run ONLY=projname CRON='8 * * * *' ./kubernetes/apply_manifest.sh ./kubernetes/manifests/devstats-hourly-sync.yaml to create a hourly sync of projname at every hour and 8 minutes.
  • Run PROJ=... PROJDB=... ICON=... ORGNAME=... ./kubernetes/apply_manifest.sh ./kubernetes/manifests/devstats-grafana.yaml to create a Grafana UI pod for a given project.
  • Run PROJ=... ./kubernetes/apply_manifest.sh ./kubernetes/manifests/devstats-service.yaml to expose Grafana via LoadBalancer service.
  • To provision all projects do: ./kubernetes/provision_them_all.sh, then wait for all devstats-provision-... pods to finish. This can take a LOT of time.
  • To setup hourly sync for all currently defined project just run: ./kubernetes/cron_them_all.sh. Do it after all initial provisioning is finished (not needed but recommended, otherwise cronjobs will wait).
  • To create all Grafanas use: ./kubernetes/grafanas_for_all.sh. Do it after all initial provisioning is finished (not needed but recommended, otherwise Grafanas will access not fully provisioned databases and some dashboards may not work) until ready).
  • To create LoadBalancer services for all Grafanas use: ./kubernetes/services_for_all.sh. Do it after grafanas are created, otherwise services will point to voids.
  • To cleanup completed pod, use: ./kubernetes/cleanup_completed_pods.sh.
  • To delete all DevStats cron jobs run: ./kubernetes/delete_devstats_cron_jobs.sh.

Helm

  • To delete and recreate cron jobs run: ./kubernetes/recreate_cron_jobs.sh. This uses Helm chart github.com/cncf/devstats-helm-lf.
  • Many other examples described in github.com/cncf/devstats-helm-lf.

New projects

See github.com/cncf/devstats-docker-images/NEW_PROJECT.md.