[Bugs] [Bug 1546732] New: Bad stat performance after client upgrade from 3.10 to 3.12

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1546732

            Bug ID: 1546732
           Summary: Bad stat performance after client upgrade from 3.10 to
                    3.12
           Product: GlusterFS
           Version: 3.12
         Component: disperse
          Assignee: bugs at gluster.org
          Reporter: ingard at jotta.no
                CC: bugs at gluster.org



Description of problem:
Our java application calls java.io.File.length() which results in a stat on the
linux fuse mounted glusterfs. After upgrading clients from 3.10 to 3.12 we got
very bad perf on these stat calls. Up to 20-30 sec. After downgrading clients
from 3.12 to 3.10 we get normal perf of around 20ms per stat.

The clients in this case mounts 5 different glusterfs. 4 of them are 3.10 with
distributed config, while the newest one is running 3.12.

The performance for stat from 3.12 clients to 3.10 distribute clusters remained
normal. The one problem that we saw was these 3.12 clients accessing the 3.12
disperse volume.

After noticing the problem we configured a standalone server as a 3.12 client
and mounted the 3.12 disperse volume thinking it could be related to the
combination of multiple serverside versions/clusters. The problem remained for
this system also.

After a while we downgraded all clients from 3.12 to 3.10 and the problem
disappeared. An interesting discovery was that the problem also disappeared
from the test system with only the 3.12 mount which was still running the 3.12
client. 


Version-Release number of selected component (if applicable):
3.10 and 3.12

How reproducible:
By upgrading clients to 3.12

Steps to Reproduce:
1.
2.
3.

Actual results:


Expected results:


Additional info:

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the assignee for the bug.