We deploy datahub to wikikube via the deployment charts repository here:
https://gerrit.wikimedia.org/r/plugins/gitiles/operations/deployment-charts/+/refs/heads/master/helmfile.d/services/datahub
When designing the system, we had intended to use a single Opensearch cluster of three nodes to support both the production and staging.
Whilst this wasn't perfect isolation between the two deployments, we had intended for the staging version to use its own indices within the cluster.
One of the values in the values-staging.yaml file is this:
global:
elasticsearch:
prefix: "staging-"The helm chart for the frontend component is supposed to set this here.
{{- with .Values.global.elasticsearch.indexPrefix }}
- name: ELASTIC_INDEX_PREFIX
value: {{ . }}
{{- end }}However, there is a typo.
This prefix is therefore not used and this means that the staging deployment shares the same cluster as the production deployment.
This means that we can't safely T329514: Upgrade Datahub to v0.10.4 because this requires a reindexing of the Opensearch indices.