Project

General

Profile

Actions
Copy link

Bug #19243

open

multimds on linode: MDSs make uneven progress on independent workloads

Added by Patrick Donnelly about 7 years ago. Updated about 5 years ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:

0%

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

Description

I have a test that just finished with 4 MDS with 400k cache object limit and 64 clients. What appears to happen in this test is that one MDS at a time makes significant progress while others make much slower or no progress. There appears to be some kind of contention which probably indicates a bug.

The experiment logs/stats are here:

mira092.front.sepia.ceph.com:/mnt/pdonnell/vault/4x8192 40000C MDS 64x4096 Client/

The caps issued graph (mds-caps.png) particularly shows the uneven progress. Also the strays graph (mds-strays.png) shows how clients for a particular MDS unlink their sandboxes roughly at the same time.

Actions
Copy link
 #1

Updated by John Spray almost 7 years ago

  • Subject changed from multimds: MDSs make uneven progress on independent workloads to multimds on linode: MDSs make uneven progress on independent workloads
Actions
Copy link
 #2

Updated by Patrick Donnelly about 5 years ago

  • Category deleted (90)
  • Labels (FS) multimds added
Actions
Copy link

Also available in: Atom PDF