Bug 762604 (GLUSTER-872)

distribute-> 2 replicates -> 4 protocol/client

shell 1:
root@ubuntu:/mnt/glusterfs# while [ true ]; do  echo 1 >> passwd; sleep 1 ; done

shell 2:
root@ubuntu:/mnt/glusterfs# tail -f passwd 
1
1

(to keep fd open)

shell 3:
kill one of the server process and bring it back up.
root@ubuntu:/home/krishna.tmp/volfiles/# ls -l /export/dir[12]
/export/dir1:
total 8
-rw-r--r-- 1 root root 431 2010-04-29 10:14 passwd

/export/dir2:
total 8
-rw-r--r-- 1 root root 311 2010-04-29 10:14 passwd
root@ubuntu:/home/krishna.tmp/volfiles/# ls -l /export/dir[12]
/export/dir1:
total 8
-rw-r--r-- 1 root root 435 2010-04-29 10:14 passwd

/export/dir2:
total 8
-rw-r--r-- 1 root root 315 2010-04-29 10:14 passwd
root@ubuntu:/home/krishna.tmp/volfiles/#

you can see that file sizes are different but still getting updated with writes.

now on shell 2, ctrl-c the tail -f command
on shell 3:
root@ubuntu:/home/krishna.tmp/volfiles/# ls -l /export/dir[12]
/export/dir1:
total 8
-rw-r--r-- 1 root root 443 2010-04-29 10:14 passwd

/export/dir2:
total 8
-rw-r--r-- 1 root root 443 2010-04-29 10:14 passwd

you can see that now it has selfhealed after close.

it is fine if it is selfhealed after close, but should not be updating the outdated copy if selfheal has not happened.

Anush tried to reproduce the issue, but couldn't reproduce. TRACE logs if you can reproduce will help.

Avati, close if this is fixed.