Project

General

Profile

Bug #41141

mds: recall capabilities more regularly when under cache pressure

Added by Patrick Donnelly 3 months ago. Updated 3 months ago.

Status:
Pending Backport
Priority:
Urgent
Category:
-
Target version:
Start date:
Due date:
% Done:

0%

Source:
Development
Tags:
Backport:
nautilus,mimic
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(FS):
MDS
Labels (FS):
Pull request ID:
Crash signature:

Description

If a client is doing a large parallel create workload, the MDS may not recall capabilities fast enough and the client capabilities grow without bound.

Two things should happen here: the MDS should more regularly recall client state. The upkeep thread currently operates every 5 seconds. Second, the MDS defaults for client recall are too conservative. A client may obtain significantly more capabilities than can be recalled.

See also: https://lists.ceph.io/hyperkitty/list/ceph-users@ceph.io/thread/AOYWQSONTFROPB4DXVYADWW7V25C3G6Z/


Related issues

Copied to fs - Backport #41467: nautilus: mds: recall capabilities more regularly when under cache pressure Resolved
Copied to fs - Backport #41468: mimic: mds: recall capabilities more regularly when under cache pressure Need More Info

History

#1 Updated by Patrick Donnelly 3 months ago

  • Status changed from New to Need Review
  • Pull request ID set to 29542

#2 Updated by Patrick Donnelly 3 months ago

  • Status changed from Need Review to Pending Backport

#3 Updated by Nathan Cutler 3 months ago

  • Copied to Backport #41467: nautilus: mds: recall capabilities more regularly when under cache pressure added

#4 Updated by Nathan Cutler 3 months ago

  • Copied to Backport #41468: mimic: mds: recall capabilities more regularly when under cache pressure added

Also available in: Atom PDF