Description of problem: Running DB workload patterns with all perf xlators enabled as default has resulted into some inconsistency issues. Based on the internal testing done by Elvir Kuric there're certain set of perf xlators which need to be turned off to get these types of workload supported by Gluster. The proposal is to leverage group profile infrastructure to group together all those tunables at one place so that users just need to apply the profile to the volume to use it for the data base workload. Version-Release number of selected component (if applicable): mainline
REVIEW: https://review.gluster.org/20414 (extras/group : add database workload profile) posted (#1) for review on master by Atin Mukherjee
There's some effort going on at BZ https://bugzilla.redhat.com/show_bug.cgi?id=1512691 to address the current gaps in the performance xlators where these issues are seen. Once those problems are addressed, the group profile may not be no longer required.
This is what Elko updated: Environment Openshift Container Platform - 3.7 Openshift Container Platform - 3.9 Red Hat Gluster Storage-3.X Issue The pgbench perf benchmark for PGSQL does not complete Issue with open-behind and write-behind xlators in Red Hat Gluster Storage when gluster-file is being used for containerized DB workload. Resolution Adapt performance translators per below recommendation in Red Hat Gluster Storage / CNS Follow the steps below to adapt performance translators Note : In case of Container Native Storage (CNS), below commands should be be executed inside one CNS pod # gluster volume set VOLUME performance.open-behind off # gluster volume set VOLUME performance.write-behind off # gluster volume set VOLUME performance.stat-prefetch off # gluster volume set VOLUME performance.quick-read off # gluster volume set VOLUME performance.strict-o-direct on # gluster volume set VOLUME performance.read-ahead off # gluster volume set VOLUME performance.io-cache off # gluster volume set VOLUME performance.readdir-ahead off
COMMIT: https://review.gluster.org/20414 committed in master by "Niels de Vos" <ndevos> with a commit message- extras/group : add database workload profile Running DB workload patterns with all perf xlators enabled as default has resulted into some inconsistency issues. Based on the internal testing done by Elko Kuric (ekuric) there're certain set of perf xlators which need to be turned off to get these types of workload supported by Gluster. The proposal is to leverage group profile infrastructure to group together all those tunables at one place so that users just need to apply the profile to the volume to use it for the data base workload. Credits : Elko Kuric (ekuric) Change-Id: I8a50e915278ad4085b9aaa3f160a33af7c0b0444 fixes: bz#1596020 Signed-off-by: Atin Mukherjee <amukherj>
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-5.0, please open a new bug report. glusterfs-5.0 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] https://lists.gluster.org/pipermail/announce/2018-October/000115.html [2] https://www.gluster.org/pipermail/gluster-users/