[Gluster-users] IO stall while 2nd brick is syncing.

[R.I.Pienaar]

Hello,

I have 2 replicated bricks setup like:

glusterfs-volgen --name glstore1 --raid 1 glbrick1:/srv/gluster glbrick2:/srv/gluster

I had to take the 2nd brick out of the storage for maintenance, on bringing it back I am doing the ls -lR resync.  

During this period IO on all the connected clients more or less stalls, any directory stats or even file reads take up to minutes to respond and this more or less destroys the storage usability.

Based on the recommendation on the irc channel I attempted to remove iocache and quickread replacing quickread with readahead on the writebehind cache.  This had no noticable effect.

Any help with debugging this would be appreciated, I am on 3.0.5 installed from official RPMs.  The files on the storage are all quite small in the 10KB region but I do have over 2 000 000 of them on the storage.

-- 
R.I.Pienaar