[Gluster-users] mv not atomic in glusterfs
Hi all, I'm using a distributed-replicated volume and it seems that mv is not atomic with a volume mounted using GlusterFS ( but works well with NFS) Volume Name: test Type: Distributed-Replicate Status: Started Number of Bricks: 2 x 2 = 4 Transport-type: tcp Bricks: Brick1: ylal3540:/users3/test Brick2: ylal3520:/users3/test Brick3: ylal3530:/users3/test Brick4: ylal3510:/users3/test Options Reconfigured: nfs.port: 2049 performance.cache-size: 6GB performance.cache-refresh-timeout: 10 performance.cache-min-file-size: 1kB performance.cache-max-file-size: 4GB network.ping-timeout: 10 features.quota: off features.quota-timeout: 60 mount -t glusterfs ylal3530:/test /users/glusterfs_mnt/ # while :; do echo test /users/glusterfs_mnt/tmp/atomic1 ;mv /users/glusterfs_mnt/tmp/atomic1 /users/glusterfs_mnt/tmp/atomic2;done # while :; do cat /users/glusterfs_mnt/tmp/atomic2 1/dev/null;done cat: /users/glusterfs_mnt/tmp/atomic2: No such file or directory cat: /users/glusterfs_mnt/tmp/atomic2: No such file or directory cat: /users/glusterfs_mnt/tmp/atomic2: No such file or directory cat: /users/glusterfs_mnt/tmp/atomic2: No such file or directory cat: /users/glusterfs_mnt/tmp/atomic2: No such file or directory cat: /users/glusterfs_mnt/tmp/atomic2: No such file or directory ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
Re: [Gluster-users] 'remove-brick' is removing more bytes than are in the brick(?)
Hi Harry, Thanks for taking time and testing 3.3.0qa42. So this was a big improvement over the previous trial. the only glitches were the 120 failures (which mean...?) You can open log file and search for the reason for the failure. (/var/lib/glusterfs/gli-rebalance.log). Currently there may be more logs for you to search, but do a grep on '\ E\ ' to get the relevant logs. and the directory skeleton left on the removed brick which may be a feature..? It is the current design limitation. We don't deal with anything directory in rebalance and remove-brick. Both these operations are on files. So it seems to have been fixed in qa42. Thanks for confirming. Regards, Amar ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
[Gluster-users] gluster volume replace-brick ... status breaks when executed on multiple nodes
I've executed gluster volume replace-brick ... status on multiple peers at the same time, which resulted in quite an interesting breakage. It's no longer possible to pause/abort/status/start the replace-brick operation. Please advise. I'm running glusterfs 3.2.6. root@ca2:~# gluster volume replace-brick sites ca1-int:/data/glusterfs ca2-int:/data/ca1 status replace-brick status unknown root@ca2:~# gluster volume replace-brick sites ca1-int:/data/glusterfs ca2-int:/data/ca1 pause replace-brick pause failed root@ca2:~# gluster volume replace-brick sites ca1-int:/data/glusterfs ca2-int:/data/ca1 abort replace-brick abort failed root@ca2:~# gluster volume replace-brick sites ca1-int:/data/glusterfs ca2-int:/data/ca1 start replace-brick failed to start -- Tomasz Chmielewski http://www.ptraveler.com ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
[Gluster-users] Questions regarding rearrangements in the gluster file backend
Hello, since we need to restructure our gluster infrastructure a wee bit, I'd need to use disks twice, aka two logical bricks on one physical disk. Current layout, mountpoint = brick: /srv/disk1/volumecontents /srv/disk2/volumecontents /srv/disk3/volumecontents /srv/disk4/volumecontents What my plan is currently: mount /mnt/disk1 mkdir /mnt/disk1/disk1 mv /mnt/disk1/* /mnt/disk1/disk1 mount --bind /mnt/disk1/disk1 /srv/disk1 [repeat...] so that the final layout looks something like that: /mnt/disk1/disk1 - /srv/disk1 /mnt/disk2/disk2 - /srv/disk2 /mnt/disk3/disk3 - /srv/disk3 /mnt/disk4/disk4 - /srv/disk4 /mnt/disk1/disk5 - /srv/disk5 *new /mnt/disk2/disk6 - /srv/disk6 *new /mnt/disk3/disk7 - /srv/disk7 *new /mnt/disk4/disk8 - /srv/disk8 *new Now, my questions: * can I do that without destroying/altering/... relevant dht information/the distribution of the files? The pathnames of what was there before stay the same for gluster. * A mv should carry the extended attributes, shouldn't it? * Can I use a symlink instead of a bind mount? Thank you for any insight on this, Philip ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
