]> git-server-git.apps.pok.os.sepia.ceph.com Git - ceph.git/commit
mon [stretch-mode]: Allow a max bucket weight diff threshold 67790/head
authorKamoltat Sirivadhna <ksirivad@redhat.com>
Tue, 9 Dec 2025 21:00:38 +0000 (21:00 +0000)
committerKamoltat (Junior) Sirivadhna <ksirivad@redhat.com>
Tue, 17 Mar 2026 21:36:20 +0000 (21:36 +0000)
commit5a94f78b43bbda036890b54ae4e114f0e0038aca
tree48deca012b2b0c3a753e7c622f5a37fbca56ed95
parent70f8415b300f041766fa27faf7d5472699e32388
mon [stretch-mode]: Allow a max bucket weight diff threshold

Problem:
Users ran into a problem where the crush bucket
weight different check in stretch mode is too strict, e.g.,
one of the disk that is added to one of the node had slight variation
in the capacity and this caused ceph to fail from enabling the stretch
cluster because crush weight is not balanced. The difference was very small.

Solution:
- Introducing: mon_stretch_max_bucket_weight_delta in mon.yaml.in
  this config var is default to 0.1 and is used as a threshold
  to allow the difference between the two crush buckets in stretch mode
  to be no greater than 10%.
- Introducing: STRETCH_MODE_BUCKET_WEIGHT_IMBALANCE as health warnings
  when the weight delta between the two sites exceeds 10%
- Modified documentations
- Modified tests that exercises this code path

Fixes: https://tracker.ceph.com/issues/72994
Signed-off-by: Kamoltat Sirivadhna <ksirivad@redhat.com>
(cherry picked from commit d58de5174d05ad2df1f1d6771abf504b25e62c54)

Conflicts:
doc/rados/operations/health-checks.rst - Trivial Fix
PendingReleaseNotes - Remove this
Signed-off-by: Kamoltat (Junior) Sirivadhna <ksirivad@redhat.com>
doc/rados/operations/health-checks.rst
qa/standalone/mon-stretch/mon-stretch-uneven-crush-weights.sh
src/common/options/mon.yaml.in
src/mon/OSDMonitor.cc
src/osd/OSDMap.cc