[SOLVED] Root partition getting temporary full during backup job

[cglmicro]

Hi everyone.

My root "/" partition is getting from ±60% to a 91% usage during backup job to an external PBS. After the job, free space is back to normal.

Like if a temp folder was pointing to a mount on this 21Gb partition instead of the 500Gb partition (i,e,: /var/lib/vz/*').

I don't see any noticeable files in /var/log/.

Where should I start to investigate?

Thank you.

 

[Dunuin]

I would monitor "/tmp" and "/var/tmp".

 

[cglmicro]

I can't pinpoint where the file is added, but it happened when doing a backup to a PBS in ZSTD, around the line:

INFO: CT Name: pmg14.legardeur.net

INFO: including mount point rootfs ('/') in backup

INFO: starting first sync /proc/1085863/root/ to /var/tmp/vzdumptmp1941818_1003

INFO: first sync finished - transferred 6.42G bytes in 18s

INFO: suspending guest

And it's not hapenning when backup to LOCAL also in ZSTD.

 

[Dunuin]

LXCs needs to be written to a temporary location first. Here it is written to "/var/tmp/vzdumptmp1941818_1003". Then its chunked and backed up to the PBS. Maybe you want to set another "tmpdir" in the /etc/vzdump.conf if your root partition/LV/dataset is too small.

 

[cglmicro]

You are a genius, thanks !!

 

[fabian]

note that this only affects 'suspend' mode backups, so if you can avoid those you don't have the tmp space requirement and your backups likely will be faster as well

 

[cglmicro]

Ok, thanks for the specifics.

It's because I'm backing up a CT and I'm not able to SNAPSHOT even if it's the selected option for this job, it always SUSPEND the VM instead.

 

[fabian]

that's usually the case if your container has volumes which are not snapshottable (storage doesn't support snapshots, volume not managed by PVE, bind mount, ..).