Red Hat Bugzilla – Bug 819490
Quick-read got hung while geo-replication session was on
Last modified: 2015-12-01 11:45:32 EST
Description of problem: In geo-replication session between master and a slave, some 60000 files of 10B were created on the master, those were synced on the slave. Later these files on the master were deleted, and started created different files, but the state of slave was stuck to the previous state ,means , the old file which were deleted on the master are not yet deleted on the slave , even after 2 days.
From state dump of the aux mount process of the geo-rep , the quick read was hung.
Version-Release number of selected component (if applicable): 3.3.0qa38
How reproducible:Tried only once
Steps to Reproduce:
1.Start a geo-rep session between master and slave ,master(dist-rep) ans slave(dist-stripe)
2.Create 60000 of 10B file on the master and let it sync on the slave
3.Delete the all files on the master and start creating many file on the master.
4. The new files created shouldn't sync on the slave.
Created attachment 582635 [details]
Statedump of the aux mount process of the geo-rep session
Created attachment 582636 [details]
Log file the aux mount process.
Even I hit the issue when I was running dbench and doing quota enable/disable in a loop and once disabled/enabled quick-read xlator also.
Statedump of the hung client process is attached.
Created attachment 584647 [details]
statedump of the hung client
The patch has gone in and for some reason its not reflected in the comments here. Hence moving bug to QA.
While reproducing the issue i found another bug 827405 , and i was not able to reproduce the above bug.
Although i got another bug while verifying this bug , i am moving to verified.