Bug 1250809 - Enable multi-threaded epoll for glusterd process
Summary: Enable multi-threaded epoll for glusterd process
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: GlusterFS
Classification: Community
Component: glusterd
Version: 3.7.0
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: ---
Assignee: krishnan parthasarathi
QA Contact:
URL:
Whiteboard:
Depends On: 1242421
Blocks: 1241336
TreeView+ depends on / blocked
 
Reported: 2015-08-06 04:36 UTC by krishnan parthasarathi
Modified: 2015-11-03 23:06 UTC (History)
3 users (show)

Fixed In Version: glusterfs-3.7.4
Clone Of: 1242421
Environment:
Last Closed: 2015-09-09 09:39:07 UTC
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Embargoed:


Attachments (Terms of Use)

Description krishnan parthasarathi 2015-08-06 04:36:10 UTC
+++ This bug was initially created as a clone of Bug #1242421 +++

Description of problem:

glusterd uses single epoll worker thread for sending/receiving notifications. gluster commands which communicate with many daemons, e.g snapshot-create could take longer than ping-timeout secs on a volume with large no. of snapshots times out. With introduction of multi-threaded epoll support in glusterd, we could add more epoll workers to (reduce latency and) avoid spurious disconnects due to ping-timeout.

Version-Release number of selected component (if applicable):


How reproducible:


Steps to Reproduce:
1.
2.
3.

Actual results:


Expected results:


Additional info:

--- Additional comment from Anand Avati on 2015-07-13 06:04:35 EDT ---

REVIEW: http://review.gluster.org/11630 (glusterd: use 2 epoll worker threads by default) posted (#2) for review on master by Krishnan Parthasarathi (kparthas)

Comment 1 Anand Avati 2015-08-24 03:45:06 UTC
REVIEW: http://review.gluster.org/11991 (event: add dispatched flag to know if event_dispatch was called) posted (#2) for review on release-3.7 by Krishnan Parthasarathi (kparthas)

Comment 2 krishnan parthasarathi 2015-08-24 08:32:43 UTC
http://review.gluster.org/11630 introduced multi-threaded epoll for glusterd on Linux systems. This patch creates incorrect no. of worker threads than the default. This is because, multi-threaded epoll subsystem doesn't allow reconfiguring of no. of threads before event_dispatch is called. Fix in comment#1 fixes that.

tl;dr
With http://review.gluster.org/11630 and http://review.gluster.org/11991 glusterd is capable of listening for epoll(7) events from 2 threads by default.

Comment 3 krishnan parthasarathi 2015-08-28 11:04:08 UTC
http://review.gluster.org/11991 doesn't fix the problem mentioned in comment#2. It has been abandoned in favour of http://review.gluster.org/12004.

Comment 4 Kaushal 2015-09-09 09:39:07 UTC
This bug is getting closed because a release has been made available that should address the reported issue. In case the problem is still not fixed with glusterfs-3.7.4, please open a new bug report.

glusterfs-3.7.4 has been announced on the Gluster mailinglists [1], packages for several distributions should become available in the near future. Keep an eye on the Gluster Users mailinglist [2] and the update infrastructure for your distribution.

[1] http://thread.gmane.org/gmane.comp.file-systems.gluster.devel/12496
[2] http://thread.gmane.org/gmane.comp.file-systems.gluster.user


Note You need to log in before you can comment on or make changes to this bug.