Migrate staging's elasticsearch inside k8s using the ECK operator
It's a POC to evaluate the capacity of the eck-operator to manage elasticsearch instances. The POC can be done without unplugging esnode0 until the deployment is fully validated and stressed.
-
Deployed an elasticsearch cluster with 3 nodes -
Trigger a search journal client to reindex the data -
Monitor the elasticsearch cluster [1] -
Chaos monkey tests -
Increase replica to 3 (from 1) ~> operator deals with the replicas and sharding -
Upgrade operator (current version 2.13, latest 2.15) -
Restart deployment (rollout) -
Kill a pod (out of a cluster with 3 replicas) -
Kill 2 pods (out of a cluster with 3 replicas) -
...
-
-
swh/infra/ci-cd/swh-charts!530 (merged): Migrate the staging search instance using the new elasticsearch instance running in kube
Note: It's the ELK stack but the official operator is named eck-operator
[1] https://grafana.softwareheritage.org/goto/Yvb3a2VHz?orgId=1
Edited by Antoine R. Dumont