Project

General

Profile

Bug #7208

CEPH_FEATURE_CRUSH_V2 feature mismatch

Added by Mark Nelson about 10 years ago. Updated about 10 years ago.

Status:
Resolved
Priority:
High
Assignee:
Category:
-
Target version:
-
% Done:

0%

Source:
Development
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

In attempting to test kernel rbd with 0.74-553-g54d51d8-1saucy on kernel 3.11, I'm seeing:

feature set mismatch, my 4a042a42 < server's 104a042a42, missing 1000000000

Which appears to be due to CEPH_FEATURE_CRUSH_V2. According to Sage, this hasn't been merged yet, but should only be necessary for erasure coded pools.

History

#1 Updated by Ian Colle about 10 years ago

  • Project changed from rbd to Ceph

#2 Updated by Samuel Just about 10 years ago

  • Priority changed from Normal to High

#3 Updated by Sage Weil about 10 years ago

  • Status changed from New to Resolved
  • Source changed from other to Development

this was fixed when i fixed up the default crush tunables to be firefly not optimal

#4 Updated by Sage Weil about 10 years ago

workarounds:

ceph osd crush tunables bobtail

works with any kernel > 3.9. or,

ceph osd crush tunables firefly

works only with the latest (i forget if it has been released yet).

if this error is coming up for a 3.11 kernel with bobtail tunables, something else is wrong (probably a different feature bit than crush v2)

#5 Updated by Sage Weil about 10 years ago

oh, 08fa34d94e40f2f7230dec4dabad72107cdee27b removed the default erasure pool from the OSDMap which fixed this symptom. it was originally broken by f8a4001440c221de575375f8bfa84fad6f3a48cb

Also available in: Atom PDF