Description of problem: Sometimes gluster volume heal $volname info returns 0 when there are files to heal Version-Release number of selected component (if applicable): 3.3.0 from git-master a032de191ec32be363d1feedfbd839f6dbde2579 How reproducible: Rsync to gluster share, stop one server check heal info Steps to Reproduce: 1. Setup replica between two nodes 2. mount on client 3. rsync filesystem on client (in my case, nfs vers4) to the gluster mount 4. While rsyncing, stop one gluster services on ONE node ONLY with /etc/init.d/glusterd stop && pkill glusterd && pkill glusterfsd 5. On the node that is still up, run gluster heal volume $volname info several times in a row Actual results: Sometimes a long list of files to heal is returned. Sometimes it says there are no files to heal. Expected results: Always should show an accurate list of files Additional info: Ubuntu 11.04 servers on 10GB e, 2 bricks each
CHANGE: http://review.gluster.org/4043 (cluster/afr: gluster volume heal $VN info sporadically reports nothing to heal) merged in master by Anand Avati (avati)
I tried building fresh servers & I can't get this condition to repeat. Nice job! Glad to see this fixed, however I did find a new/related issue (?) if I run "volume heal data info" and gluster hasn't started outputting the results of this command I run the same command from another shell, I immediately get "operation failed" on the second shell. If I repeat this after the first command starts returning data, the second one will work. Timing issue? Should I open a new bug report for this?