We have an ElasticSearch cluster on ElastiCloud on one region and we want to migrate it to another region.
we saw there are few options of doing so, and one of them which seems to be the easiest was do it using Snapshots.
so i've created the following process:
- Create GCS snapshot repository on the Source Cluster
- Create snapshot for the source cluster on the GCS snapshot repo
- Create GCS snapshot reposirory (read only) on the new Cluster
- Sync both GCS Buckets (GCP commands)
- Restore Target cluster from the snapshot which was added to the snapshot repository
now, i've noticed this process should includes the following:
- Include global state so ILM policies will also be included on the migration (ilm indices)
- Close all migrated indices before restore
the thing is, although restore completed successfully, i've noticed after restore, all migrated indices switched to yellow (or even red) as they are being unassigned and rellocated to the nodes.
this operation can take a lot of time (20min on our case)
is this the expected behaviour, reassigning all indices?
anything we can do to improve this?