Since we can't use our typical test cluster (relforge) as it's occupied by a Semantic Search experiment (ref T413969 ), we will start our OpenSearch 2.x migration on the beta cluster. Creating this ticket to:
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | bking | T421757 ☂️ Migrate production OpenSearch clusters from 1.x-2.x ☂️ | |||
| Duplicate | bking | T421763 Migrate beta cluster to OpenSearch 2.x | |||
| Resolved | bking | T424100 No Puppet resources found on instance deployment-cirrussearch14 on project deployment-prep | |||
| Resolved | bking | T424104 No Puppet resources found on instance deployment-cirrussearch12 on project deployment-prep | |||
| Resolved | bking | T425585 Write lightweight OCI-image-based Puppet plans for beta cluster |
Upon further review, role::cirrus::beta just points back to the production profiles, so there's no real benefit to using Beta for us. We'll start in Cloudelastic instead, see T422860.
Reopening, as it seems we don't necessarily need to update the Puppet code to update the Beta cluster. I think we can override the hieradata for the OpenSearch clusters in beta using the OpenStack web UI. Will give it a try and report back with my findings shortly.
Mentioned in SAL (#wikimedia-cloud) [2026-04-21T20:47:23Z] <inflatador> updating cirrussearch hosts to Trixie/OpenSearch 2 T421763
Update before the Earth Day holiday:
I changed the hieradata, but Puppet didn't like it. It set off some alerts (see linked ticket) and I have ACK'd them all. The cluster itself is still healthy.
I also provisioned deployment-cirrussearch15.deployment-prep.eqiad1.wikimedia.cloud as part of the replacement process, but I cannot login and see what went wrong there, presumably because Puppet failed at an early stage. I'll console in and take a look after the holiday.