I have googled and tried some of the archives of the list to locate a good solution to this issue. I pre-apologize if this has been beaten to death in the past.
It seems that some sort of order or timing is a bit off when I attempt to make a dirvish backup of an LVM snapshot. I have a pre-client call that executes a shell script on the client that is as follows: #!/bin/sh FS=/dev/VolGroup00/$1 SNAP=SNAP-$1 FSS=/dev/VolGroup00/$SNAP lvcreate -L512M -s -n $SNAP $FS mount $FSS /mnt/SNAPSHOTS/$1 I have a post-client that "undoes" that work: #!/bin/sh sleep 15 FS=/dev/VolGroup00/$1 SNAP=SNAP-$1 FSS=/dev/VolGroup00/$SNAP echo "SNAPSHOT of $FS" umount $FSS lvremove -f $FSS It SEEMS (but I'm not sure) of the following A) When my dirvish server connects to the client, it attempts to cd to the target tree prior to running the pre-client. I have no idea why that would be, but it doesn't cause the rest of it to fail. B) When my dirvish server is finished, the process remains cd'd in the tree filesystem. This causes the umount of post-client script to fail, causing the lvremove to fail, leaving my client with a potentially ugly situation involving a residual LVM snapshot. My current solution is to pre-pend a "cd / ; " to the post client script but this seems to be by intent, either a bug or a feature. Does anyone have any thoughts on this? Are my assumptions erroneous? Is it something in rsync's behavior rather than dirvish? Thanks, Mykel
_______________________________________________ Dirvish mailing list [email protected] http://www.dirvish.org/mailman/listinfo/dirvish
