Project

General

Profile

Feature #57557

Ability to roll-back the enabled stretch-cluster configuration

Added by Dmitry Smirnov 3 months ago. Updated 3 months ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
Stretch Clusters
Target version:
-
% Done:

0%

Source:
Tags:
stretch-cluster
Backport:
Reviewed:
Affected Versions:
Component(RADOS):
Pull request ID:

Description

We have enabled a stretch-cluster configuration on a pre-production system with several already existing and used pools.

  1. ceph mon dump
    epoch 64
    fsid 0cd77d50-793e-4579-bad5-461b799d7620
    last_changed 2022-09-15T15:07:47.523551+0300
    created 2022-07-01T11:29:41.383919+0300
    min_mon_release 16 (pacific)
    election_strategy: 3
    stretch_mode_enabled 1
    tiebreaker_mon bkus-cnode3
    disallowed_leaders bkus-cnode3
    0: [v2:10.67.32.1:3300/0,v1:10.67.32.1:6789/0] mon.bkus-cnode1; crush_location {datacenter=k35}
    1: [v2:10.67.32.2:3300/0,v1:10.67.32.2:6789/0] mon.bkus-cnode2; crush_location {datacenter=k25}
    2: [v2:10.67.32.3:3300/0,v1:10.67.32.3:6789/0] mon.bkus-cnode3; crush_location {datacenter=k15}
    dumped monmap epoch 64
  1. ceph osd dump |grep stretch
    stretch_mode_enabled true
    stretch_bucket_count 2
    degraded_stretch_mode 0
    recovering_stretch_mode 0
    stretch_mode_bucket 8

After a series of tests we realized this configuration is not suitable for our case. I can't find a way how to fully roll-back CEPH to its original state.
I can roll-back MON epoch and get rid of stretch-cluster mode, but cannot revert OSDmap epoch due to large amount of changes made and OSD epoch versions expiration.

History

#1 Updated by Neha Ojha 3 months ago

  • Tracker changed from Support to Feature
  • Project changed from Ceph to RADOS
  • Category changed from Monitor to Stretch Clusters

As discussed in https://bugzilla.redhat.com/show_bug.cgi?id=2094016, this hasn't been implemented yet.

Also available in: Atom PDF