Project

General

Profile

Bug #38787

osd: cache tiering flush clone wrongly

Added by Zengran Zhang 7 months ago. Updated 7 months ago.

Status:
Need Review
Priority:
Normal
Assignee:
Category:
Snapshots
Target version:
Start date:
Due date:
% Done:

0%

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

Description

because cephfs file snapcontext seq may start from 1, we find that in a never snaped fs,
the flush of file will delete the head with snapcontext 0, and send copy_from with snapcontext 1,
although the copy_from will remove base object, but the key point is deletion and copy_from is not atomic.

History

#1 Updated by Patrick Donnelly 7 months ago

  • Category set to Snapshots
  • Status changed from New to Need Review
  • Assignee set to Zengran Zhang
  • Target version set to v15.0.0
  • Start date deleted (03/17/2019)
  • Source set to Development
  • Backport set to mimic
  • Pull request ID set to 27016
  • Component(RADOS) OSD added

Also available in: Atom PDF