Skip to content

[mimir][pre] - adds temporary options to assit migrate ingester disk to ssd

Nick Duffrequested to merge
nduff/mimir-pre into master

What

  • Sets querier.query-store-after to 30m, default is 12hours
  • Sets blocks-storage.tsdb.flush-blocks-on-shutdown to true for ingesters

Why

This enables us to test migrating to ssd for the ingesters in pre.

By default we only query store gateways after 12hours, where the first 12 hours of data lives on the ingester persistent disks. When we look to migrate the ingester disk to SSD we need to change this so we query the stores much sooner.

The additional flag also ensures what when we shutdown the ingesters we flush the blocks on disk to long term storage.

This enables us to shutdown a zone of ingesters, flush to disks to GCS, and replace the disk with SSD.

By reducing the query store after period, this also allows us to do 1 zone per 30minutes, rather than needing to wait 12 hours between zones.

Merge request reports