0 backup files 250GB+ HD is still taken ?!

[chudak]

Hello all,

I noticed that my HD space (using zfs single disk) is shrinking and did a test.

Removed all backups =>





But sill see 250GB+ space taken +.



Whats is going on and how to free the HD? I was hoping Prune and GC would do that, but it does not seem to do so.

Thx

 

[aaron]

GC will only remove chunks once they are older than 24 hours. This has technical reasons, see the second note here: https://pbs.proxmox.com/docs/backup-client.html#garbage-collection

 

[chudak]

GC will only remove chunks once they are older than 24 hours. This has technical reasons, see the second note here: https://pbs.proxmox.com/docs/backup-client.html#garbage-collection

Thx @aaron

I kept pbs server with no backups and still see the space used



So quotations :

1 - will it be ever freed ?
2 - is it OK ?
3 - how to clean it up?

Thx

 

[chudak]

I see now.
Ran Prune & CC and now see the spaced free.
That's cool, but it's not automatic ?!

 

[aaron]

I see now.
Ran Prune & CC and now see the spaced free.
That's cool, but it's not automatic ?!

You can define schedules for prune and GC and verification jobs for each datastore. Be careful with verification schedules because they will need to read the whole datastore and thus will take quite some time.

 

[chudak]

You can define schedules for prune and GC and verification jobs for each datastore. Be careful with verification schedules because they will need to read the whole datastore and thus will take quite some time.

Still removed 3 out of 4 snapshots and the drive does not show space size changed.

???

 

[aaron]

It is possible that the chunks were not older than 24 hours (see 2nd note https://pbs.proxmox.com/docs/backup-client.html#garbage-collection) or that they are used by other backups as well?

 

[chudak]

It is possible that the chunks were not older than 24 hours (see 2nd note https://pbs.proxmox.com/docs/backup-client.html#garbage-collection) or that they are used by other backups as well?

I am positive.
See this:



Initial backup was ~<200GB for 6-7 VMs

Yesterday 24+ h I removed all but one snapshot.


Ran GC
....
2021-02-17T09:32:33-08:00: Original data usage: 120.00 GiB
2021-02-17T09:32:33-08:00: On-Disk usage: 95.98 GiB (79.98%)
2021-02-17T09:32:33-08:00: On-Disk chunks: 30094
2021-02-17T09:32:33-08:00: Deduplication factor: 1.25
2021-02-17T09:32:33-08:00: Average chunk size: 3.27 MiB
2021-02-17T09:32:33-08:00: TASK OK

why does the graph show 300GB+ taken?

And the dashboard shows:



df -h => PBS_ZFS 899G 98G 802G 11% /mnt/datastore/PBS_ZFS

It seems that the top graph (datastore Summary) does not show the correct value.

Agree?

 

[aaron]

It does look a bit off, but two questions: What if you set the time interval for the graph to one hour? (top right)
Since this is a ZFS dataset, what is the output of zfs list?

I don't see that effect in my personal productive PBS installation :/

 

[chudak]

It does look a bit off, but two questions: What if you set the time interval for the graph to one hour? (top right)
Since this is a ZFS dataset, what is the output of zfs list?

I don't see that effect in my personal productive PBS installation :/

What is interesting that the summary graph went down to 105.05GB but still not exact number as in the dashboard.
Is graph updates delayed?!

zfs list
NAME USED AVAIL REFER MOUNTPOINT
PBS_ZFS 272G 627G 272G /mnt/datastore/PBS_ZFS

(272G already after new backup ran, so the expected number was 98G)

And the interval was set for "monthly average".

This is how it looks today (again after one more backup was added):