I have a server on which I am implementing a backup scheme using LVM snapshots. The
logical volumes that will be snapshotted contain a proprietary database format, not a
conventional Linux filesystem. I point a /dev/raw/ device at the logical volumes
through which normal I/O is accomplished. A few questions have come up as I implement
the planned design of the backup procedures.
The logical volumes to be backed up are relatively small in their size (approx 2GB to
3GB). The number of volumes that must be *concurrently* snapshotted is relatively
large (approx 64+). By "concurrently," I mean that each of the volumes to be backed
up will have an "lvcreate --snapshot" command including a generous "-l 25%ORIGIN"
option issued for it before the backups begin to get the best consistent view of the
entire database. (My code checks the percent utilized of the snapshot before deletion
and I hope to tune that 25%ORIGIN downward depending upon experience. There is over
1TB available free space in the VG.)
When I initially created the volume group, I specified a metadatasize of 32M.
The simplified basic plan is to
1) quiesce writes to each device in the database
2) snapshot each logical volume in the set to be backed up
3) resume write capability to the original volumes in the database
4) create a /dev/raw/ device pointing at the snapshot logical block device
5) serially backup each /dev/raw/<snapshot> device using a
proprietary backup program to read it
6) delete each /dev/raw/<snapshot> and /dev/<volume-group>/<snapshot>
as its backup is completed.
Testing thus far has been successful, but I have some questions based on observations
of the testing.
* Do I need to be concerned about the effects on the metadata of snapshotting so many
devices concurrently? As it adds and deletes each snapshot device the space in the
metadata is recovered, correct? Current backups of the metadata are less than 30K bytes.
* Do I need to be concerned with the number of archive files that I see are being
created in the /etc/lvm/archive/ directory? Will they be trimmed over time or should
I implement something to remove older archive files?
* Anything else?
--
Gary Eheman
_______________________________________________
linux-lvm mailing list
linux-lvm@redhat.com
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
