Description of problem: When glusterfs-master performance is compared with 3.12.15 release (ie, last of 3.12 series), we are finding a lot of regression on the master branch. Version-Release number of selected component (if applicable): master How reproducible: 100% Steps to Reproduce: 1. run the gbench tests. 2. 3.
REVIEW: https://review.gluster.org/22107 (features/sdfs: disable by default) posted (#1) for review on master by Amar Tumballi
REVIEW: https://review.gluster.org/22107 (features/sdfs: disable by default) merged (#3) on master by Atin Mukherjee
It might be worth to open a different bugzilla to track the perf regression for sdfs only considering this bug is used as a tracker.
REVIEW: https://review.gluster.org/22120 (inode: Reduce work load of inode_table->lock section) posted (#6) for review on master by Amar Tumballi
REVIEW: https://review.gluster.org/22156 (inode: granular locking) posted (#1) for review on master by Amar Tumballi
REVIEW: https://review.gluster.org/22183 (inode: create inode outside locked region) posted (#1) for review on master by Amar Tumballi
REVIEW: https://review.gluster.org/22184 (inode: make critical section smaller) posted (#1) for review on master by Amar Tumballi
REVIEW: https://review.gluster.org/22185 (inode: dentry_destroy outside of dentry_unset) posted (#1) for review on master by Amar Tumballi
REVIEW: https://review.gluster.org/22186 (inode: don't take lock on whole table during ref/unref) posted (#1) for review on master by Amar Tumballi
REVIEW: https://review.gluster.org/22188 (inode: do only required checks inside critical section.) posted (#1) for review on master by Amar Tumballi
REVIEW: https://review.gluster.org/22183 (inode: create inode outside locked region) merged (#4) on master by Amar Tumballi
REVIEW: https://review.gluster.org/22184 (inode: make critical section smaller) merged (#10) on master by Amar Tumballi
REVIEW: https://review.gluster.org/22242 (inode: reduce inode-path execution time) posted (#1) for review on master by Amar Tumballi
REVIEW: https://review.gluster.org/22243 (inode: handle list management outside of ref/unref code) posted (#1) for review on master by Amar Tumballi
Has git bisect been used to narrow down the patches that have caused the regression? The inode code has not changed in a long time so this is unlikely to be the cause of the slowdown.
> Has git bisect been used to narrow down the patches that have caused the regression? The inode code has not changed in a long time so this is unlikely to be the cause of the slowdown. Ack, inode code is surely not the reason. The code/features identified as reasons for some of the regressions were: * no-root-squash PID for mkdir-layout set code in DHT (for mkdir) * gfid2path xattr setting (for rename) * ctime setting (for rmdir and few other entry ops, which seemed minor).
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-6.0, please open a new bug report. glusterfs-6.0 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/2019-March/000120.html [2] https://www.gluster.org/pipermail/gluster-users/
REVISION POSTED: https://review.gluster.org/22242 (inode: reduce inode-path execution time) posted (#3) for review on master by Amar Tumballi