Project

General

Profile

Backport #36640

luminous: Unable to recover from ENOSPC in BlueFS

Added by Patrick Donnelly 12 months ago. Updated 2 months ago.

Status:
New
Priority:
Normal
Assignee:
Target version:
-
Release:
luminous
Crash signature:

Related issues

Copied from bluestore - Bug #36268: Unable to recover from ENOSPC in BlueFS Resolved 10/01/2018

History

#1 Updated by Patrick Donnelly 12 months ago

  • Copied from Bug #36268: Unable to recover from ENOSPC in BlueFS added

#2 Updated by Nathan Cutler 11 months ago

  • Status changed from New to Need More Info
  • Assignee set to Igor Fedotov

Igor writes in the parent issue: "In fact previously mentioned PR is just a workaround to be able to manually fix the issue.
Working on the actual solution to fix BlueFS allocation strategy."

#3 Updated by Nathan Cutler 8 months ago

  • Status changed from Need More Info to New

#4 Updated by Prasad Krishnan 2 months ago

It appeared to me that increasing "bluefs_min_log_runway" config option to a really high value is one way to prevent this problem from occuring on OSDs where the fix (https://github.com/ceph/ceph/pull/25132/commits) isn't available, since the write operations would get prevented early (in case of low-space) and compaction can happen without trampling on the last available free-space.

Can someone comment on whether this looks like a good workaround to prevent the condition?

Also available in: Atom PDF