Description of problem: Recently (in rhgs-3.4.0 cycle) to fix ping-timeout issue we introduced a thread to process all requests inbound to Glusterfs program[1]. With bricks multiplexed and all bricks served as part of same program, huge traffic of requests can be expected. This thread is likely to be a bottleneck. However, there is some evidence that this thread can be a bottleneck (though with NVMe backend)[2] [1] https://review.gluster.org/17105 [2] https://bugzilla.redhat.com/show_bug.cgi?id=1467614#c51 Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
Sunil - We need to have a release candidate tag for this considering the fix was reverted due to the revert of the ping timer expiry patches.
This issue does not apply to RHGS 3.4.0 since the patches inherited from the rebase to 3.12.2 upstream have been reverted in RHGS 3.4.0. Closing as NOTABUG.