Uploaded image for project: 'OpenShift Logging'
  1. OpenShift Logging
  2. LOG-8321

Add ability to scale-down the number of Ingesters

XMLWordPrintable

    • Icon: Task Task
    • Resolution: Unresolved
    • Icon: Normal Normal
    • None
    • Logging 6.5.0
    • Log Storage
    • None
    • Product / Portfolio Work
    • False
    • Hide

      None

      Show
      None
    • False
    • Not Selected
    • NEW
    • NEW
    • If Release Note Needed, Set a Value
    • Logging - Sprint 283

      Context

      Loki Ingesters hold a significant buffer of unflushed data in RAM and participate in a strict quorum-based hash ring. Because of this we cannot simply scale down it's StatefulSet as the application will only get a SIGTERM from Kubernetes. Instead what should happen is the following:

      • We should set the ingester to LEAVING in the memberlist righ
      • We should call /ingester/shutdown?flush=true to force a flush of the ingester
      • Wait for the flush to complete
      • Lower the number of Ingester Replicas in the StatefulSet

      With this it's clear that the operator will need to interact with the ingester instance. It's not something we currently do so that will also be something to work on

      Acceptance Criteria

      • We should have a mechanism to interact with the Loki components from the operator code
      • We should be able to increase & decrease the number of ingesters without losing data

              Unassigned Unassigned
              jmarcal@redhat.com Joao Marcal
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated: