Repository navigation
|
We have a backup system that relies on a central server that pulls snapshots from clients and then periodically pushes the entire set to an offsite system. When pulling the snapshots, I do PULL command for each client (runs multiple times per day): And then the PUSH command (runs once per week): Permissions for the delegated user on both ends are: For the above, I would expect a maxium of 10 snapshots per filesystem but instead, on each client there are up to 29 snapshots of various client filesystems, which is reflected 1 <-> 1 on the backup system. Example list of snapshot numbers for one host: However yesterday I received an alert that our offsite system was running low on disk, which should not happen. Upon investigation, I was horrified to find found that no snapshot thinning was happening at all. Each FS having up to 3000 snapshots. Example for the same system: What is happening here? What am I doing wrong? |
Replies: 4 comments 5 replies
|
As a side note, I did remove "--no-holds" and "--remove-incompatible" which were accidentally left over from testing. I then did another run for each with no changes observed. |
|
SOLVED! Thinning is primarily based on snapshot names. The snapshot name format is We use different tags/properties for doing the local backups vs offsite push. For example: The offsite backup was looking for the wrong snapshots to thin. By adding I am now watching as our backup system thins out thousands upon thousands of needless datasets. This simple-but-critical misinterpretation led to some pretty drastic results. Please add a note in the FAQ with dancing ladies and flashing lights. It may save someone a significant amount of angst. |
|
you didnt mention --snapshot-format in your initial question? |
|
@psy0rz No, the command is exact copy/paste. As I stated before (see above), I did not use the `--snapshot-format` parameter at all, being happy with the default. The problem is that the variables for `$BACKUP_NAME` are assigned different values at execution during the PULL and PUSH. I did not realize this mattered. The snapshot filenames don't get renamed when pushed to the offsite server but the autobackup command is looking to thin snapshots on the offsite target with the same name as the currently running backup and therefore ignore all the snapshots being pushed. I updated the PUSH command to **now** use `--snapshot-format=originalbackupname-%Y%m%d%H%M%S`
On Apr 14, 2025, at 03:05, DatuX ***@***.***> wrote:
so you didn't post the actual command that gave you trouble, and left out the --snapshot-format??
why would you do that
—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>
|
SOLVED!
Thinning is primarily based on snapshot names. The snapshot name format is
{}-%Y%m%d%H%M%S. I failed to fully consider that{}means snapshot name and the snapshot names I am concerend with are those that were made from the origin systems, as we are not taking new snapshots, just shipping the existing snapshots that were taken using the original name.We use different tags/properties for doing the local backups vs offsite push. For example:
The offsite backup was looking for the wrong snapshots to thin. By adding
--snapshot-format local-%Y%m%d%H%M%Swe then get: