Re: [Gluster-users] Inconsistent volume
On 07/26/2010 04:37 PM, Andy Pace wrote: I too would like to know how to "sync up" a replicated pair of bricks. Right now I've got a slight difference between the 2... Scale-n-defrag.sh didn't do much either. Looking forward to some help :) 13182120616 139057220 12362648984 2% /export Vs 13181705324 139057208 12362233500 2% /export Granted, it's a very small amount (and the total availalble is slightly different), but the amount used should be the same, no? -Original Message- From: gluster-users-boun...@gluster.org [mailto:gluster-users-boun...@gluster.org] On Behalf Of Steve Wilson Sent: Monday, July 26, 2010 3:35 PM To: gluster-users@gluster.org Subject: [Gluster-users] Inconsistent volume I have a volume that is distributed and replicated. While deleting a directory structure on the mounted volume, I also restarted the GlusterFS daemon on one of the replicated servers. After the "rm -rf" command completed, it complained that it couldn't delete a directory because it wasn't empty. But from the perspective of the mounted volume it appeared empty. Looking at the individual bricks, though, I could see that there were files remaining in this directory. My question: what is the proper way to correct this problem and bring the volume back to a consistent state? I've tried using the "ls -alR" command to force a self-heal but for some reason this always causes the volume to become unresponsive from any client after 10 minutes or so. Some clients/servers are running version 3.0.4 while the others are running 3.0.5. Thanks! Steve -- Steven M. Wilson, Systems and Network Manager Markey Center for Structural Biology Purdue University (765) 496-1946 ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users All, Does anyone have any hints on how to proceed on this? I do now have all three servers running at version 3.0.5. Running a "du -bs" on each of the bricks shows the following: Replication pair #1: 693999264161/gluster/jiang-scratch-1a 693998682461/gluster/jiang-scratch-1b Replication pair #2: 231056270208/gluster/jiang-scratch-2a 231049560706/gluster/jiang-scratch-2b Replication pair #3: 228227559462/gluster/jiang-scratch-3a 228839698590/gluster/jiang-scratch-3b Is there something I can do manually (and safely) that will bring my volume back to a consistent state? Thanks, Steve -- Steven M. Wilson, Systems and Network Manager Markey Center for Structural Biology Purdue University (765) 496-1946 ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
Re: [Gluster-users] Inconsistent volume
I too would like to know how to "sync up" a replicated pair of bricks. Right now I've got a slight difference between the 2... Scale-n-defrag.sh didn't do much either. Looking forward to some help :) 13182120616 139057220 12362648984 2% /export Vs 13181705324 139057208 12362233500 2% /export Granted, it's a very small amount (and the total availalble is slightly different), but the amount used should be the same, no? -Original Message- From: gluster-users-boun...@gluster.org [mailto:gluster-users-boun...@gluster.org] On Behalf Of Steve Wilson Sent: Monday, July 26, 2010 3:35 PM To: gluster-users@gluster.org Subject: [Gluster-users] Inconsistent volume I have a volume that is distributed and replicated. While deleting a directory structure on the mounted volume, I also restarted the GlusterFS daemon on one of the replicated servers. After the "rm -rf" command completed, it complained that it couldn't delete a directory because it wasn't empty. But from the perspective of the mounted volume it appeared empty. Looking at the individual bricks, though, I could see that there were files remaining in this directory. My question: what is the proper way to correct this problem and bring the volume back to a consistent state? I've tried using the "ls -alR" command to force a self-heal but for some reason this always causes the volume to become unresponsive from any client after 10 minutes or so. Some clients/servers are running version 3.0.4 while the others are running 3.0.5. Thanks! Steve -- Steven M. Wilson, Systems and Network Manager Markey Center for Structural Biology Purdue University (765) 496-1946 ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
[Gluster-users] Inconsistent volume
I have a volume that is distributed and replicated. While deleting a directory structure on the mounted volume, I also restarted the GlusterFS daemon on one of the replicated servers. After the "rm -rf" command completed, it complained that it couldn't delete a directory because it wasn't empty. But from the perspective of the mounted volume it appeared empty. Looking at the individual bricks, though, I could see that there were files remaining in this directory. My question: what is the proper way to correct this problem and bring the volume back to a consistent state? I've tried using the "ls -alR" command to force a self-heal but for some reason this always causes the volume to become unresponsive from any client after 10 minutes or so. Some clients/servers are running version 3.0.4 while the others are running 3.0.5. Thanks! Steve -- Steven M. Wilson, Systems and Network Manager Markey Center for Structural Biology Purdue University (765) 496-1946 ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users