Description of problem: In one of the performance testing effort, we found that below options help improve the overall performance numbers of gluster-block workload: performance.io-thread-count 16 -> 32 performance.high-prio-threads 16 -> 32 performance.normal-prio-threads 16 -> 32 performance.low-prio-threads 16 -> 32 performance.least-prio-threads 1 -> 4 client.event-threads 2 -> 8 server.event-threads 1 -> 8 performance.client-io-threads off -> on I was thinking its good idea to add this changes to current block profile, as we have data showing improvements in perf results. Also, its good idea to re-validate the existing gluster-block profile options in /var/lib/glusterd/groups/gluster-block
Hello Elvir, Would you able to help us test the bug for the perf improvements if we accept this for ocs 3.11.4 ? Thanks kasturi
Amar & Atin, The fix for this bug has to come from various perf (translator) experts within the core gluster team. We need collaboration from RHGS experts to help tune these volume options for better block performance. I see Amar had assigned it to me, but ideally speaking we need various perf experts to take a look and add their suggestions/opinions on this. Hope that makesense. Thanks!
> Hope that makesense. Yes it does. I would get started with few basic things we already know in the profile info, and tuning further if required. My idea is, the basic tuning should get in to 3.5.0 release, and there can be further improvements which can get into later releases. Adding other perf experts in the bug as CC. Team, we surely need to work on this collaboratively.
PR: https://review.gluster.org/#/c/glusterfs/+/22963/
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. https://access.redhat.com/errata/RHEA-2019:3249