Red Hat Bugzilla – Bug 818153
[glusterfs-3.3.0qa39] - remove-brick commit force doesn't work as expected for a replicate volume.
Last modified: 2016-05-31 21:56:02 EDT
Description of problem: I have a 3 node replicate volume. When I issue remove-brick to to make it 2 node replica, commit force just throws the usage message. remove-brick commit force should have removed the brick. But just remove-brick without any option works. Version-Release number of selected component (if applicable): glusterfs-3.3.0qa39 How reproducible: Consistent Steps to Reproduce: 1. Create and start a 3 node replicate volume. And create some data on the mountpoint. 2. Now run remove-brick <volname> <brick3> start 3. Then run remove-brick <volname> <brick3> status 4. remove-brick <volname> <brick3> commit 5. remove-brick <volname> <brick> commit force Actual results: I have pasted below the sequence of commands I executed and it's output [root@QA-24 ~]# gluster v i Volume Name: hosdu Type: Replicate Volume ID: ca61fb62-5061-46db-bf29-be3561053874 Status: Started Number of Bricks: 1 x 3 = 3 Transport-type: tcp Bricks: Brick1: 172.17.251.63:/data/bricks/hosdu_brick1 Brick2: 172.17.251.66:/data/bricks/hosdu_brick2 Brick3: 172.17.251.65:/data/bricks/hosdu_brick3 Options Reconfigured: diagnostics.count-fop-hits: on diagnostics.latency-measurement: on [root@QA-24 ~]# gluster v remove-brick hosdu 172.17.251.65:/data/bricks/hosdu_brick3 start Remove Brick start successful [root@QA-24 ~]# gluster v remove-brick hosdu 172.17.251.65:/data/bricks/hosdu_brick3 status Volume hosdu is not a distribute volume or contains only 1 brick. Not performing rebalance [root@QA-24 ~]# gluster v remove-brick hosdu 172.17.251.65:/data/bricks/hosdu_brick3 commit Removing brick(s) can result in data loss. Do you want to Continue? (y/n) y use 'force' option as migration is in progress [root@QA-24 ~]# gluster v remove-brick hosdu 172.17.251.65:/data/bricks/hosdu_brick3 commit force wrong brick type: commit, use <HOSTNAME>:<export-dir-abs-path> Usage: volume remove-brick <VOLNAME> [replica <COUNT>] <BRICK> ... {start|stop|status|commit|force} I saw the rebalance process started in the <host3>. Expected results: For a replicate volume rebalance shouldn't be started. So for remove-brick commit should not throw that error and commit force should actually commit the brick.
Vishwa, This is the expected behavior. For the given use case, where you want to convert a replica(3) volume to replica(2) volume, please provide the command gluster volume remove-brick <VOLNAME> replica 2 <BRICK> (and no need of start, status or commit). If it doesn't work, re-open the bug.