Project

General

Profile

Actions
Copy link

Bug #17408

closed

Possible un-needed wait on rstats when listing dir?

Added by John Spray over 7 years ago. Updated almost 7 years ago.

Status:
Can't reproduce
Priority:
Normal
Assignee:
-
Category:
Performance/Resource Usage
Target version:
-
% Done:

0%

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

Description

When a client does "ls -l" on a dir where other clients are doing lots of writes, it can't see an up to date size without waiting for the other clients to flush their writes. That part is unavoidable within our rule that clients only see one another's writes once they're persistent.

However, it is possible that we are waiting too long for the rstats on the directory itself. Let's check exactly what we are doing in the case of the ls -l on a folder with lots of files busy being written from another client.

Related to: https://bugzilla.redhat.com/show_bug.cgi?id=1378142

Actions
Copy link
 #1

Updated by Greg Farnum almost 7 years ago

  • Assignee deleted (Greg Farnum)
Actions
Copy link
 #2

Updated by Zheng Yan almost 7 years ago

  • Status changed from New to Can't reproduce

can't reproduce. probably fixed by 

commit d463107473382170c07d9250bb7ace0e5a2a7de2
Author: Yan, Zheng <zyan@redhat.com>
Date:   Fri May 6 19:07:07 2016 +0800

    mds: finish lock waiters in the same order that they were added.

    Current code first processes lock waiters who have smaller wait mask.
    Lock waiters who have large wait mask can starve if client keeps
    sending requests that add waiter with small mask.

    Signed-off-by: Yan, Zheng <zyan@redhat.com>

Actions
Copy link

Also available in: Atom PDF