+++ This bug was initially created as a clone of Bug #1368451 +++
Description of problem:
We need a way to increase parallelism in number of heals that can happen in disperse volumes as well. This bug tracks that RFE
Version-Release number of selected component (if applicable):
Steps to Reproduce:
--- Additional comment from Vijay Bellur on 2016-08-19 10:06:55 EDT ---
REVIEW: http://review.gluster.org/15083 (cluster/ec: Do multi-threaded self-heal) posted (#2) for review on master by Pranith Kumar Karampuri (firstname.lastname@example.org)
--- Additional comment from Worker Ant on 2016-08-24 13:21:08 EDT ---
REVIEW: http://review.gluster.org/15083 (cluster/ec: Do multi-threaded self-heal) posted (#3) for review on master by Pranith Kumar Karampuri (email@example.com)
--- Additional comment from Worker Ant on 2016-08-24 18:24:26 EDT ---
COMMIT: http://review.gluster.org/15083 committed in master by Pranith Kumar Karampuri (firstname.lastname@example.org)
Author: Pranith Kumar K <email@example.com>
Date: Thu Aug 4 00:41:16 2016 +0530
cluster/ec: Do multi-threaded self-heal
Signed-off-by: Pranith Kumar K <firstname.lastname@example.org>
Smoke: Gluster Build System <email@example.com>
Reviewed-by: Ashish Pandey <firstname.lastname@example.org>
NetBSD-regression: NetBSD Build System <email@example.com>
CentOS-regression: Gluster Build System <firstname.lastname@example.org>
1) Create 4+2 volume.
2) Kill a brick process from the server.
3) Start IO using dd with creating multiple small files in thousands
4) Open the log files on all the bricks
5) start the brick force by gluster volume <vol-name> start force
6) logs should show healing in parallel
Check for the threads in TOP command.
Heals should be done. from source brick to empty brick.
Threads should work parallel and heal files
Summary:- Healing with single thread and Multithread.
Note turn of metadata , entry and data heal and self heal daemon.
1) Create single 4+2 volume on two setups.
2) Change the "cluster.shd-max-threads" to default on one setup and 16 to second setup.
3) Kill one of the brick process and create same mulitple files and folders on the two volumes
4) Now start the metadata , entry and data heal and self heal daemon and the brick process.
5) Calculate the time for both the volumes, Mulithreaded volume should heal faster.
6) check logs for heal progress.
Mulitthreaded heal should be done faster.
No heal should fail.
verfied mtsh for ec volume and I see that parallel heals are happening to improve performance.
Moving to verified.
Will raise new bugs if any on this seperately
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.