Slow loading datastore content or timed out

[egoistdream]

Hi,


I see that if you have may snapshots on the datastore when you want to see the content is loading very very slow, and each time when you click on another datastore and come back you need to wait again the same time, also on the proxmox host which is using the backup server the loading of the content is quite slow and some time it getting connection timed out.

If you click on one vm and go to the backup for this vm the loading of the backup is faster but if you click on the entire datastore and want to print all the backup files most of the time you will get connection timedout and this happens only if you have a lot of backup.

When I had 6000 snapshots there was no way to list the backups files and I was needed to purge some backups, now I have about 3000 it is faster but not enoght.

Is there any way to extend the temeout or is there any way to index the files in the future in order to speedup the listing?


Regards

 

[itNGO]

same here.... does not work very will with many backup sets....

 

[t.lamprecht]

Yes this has room for improvement on setups with many snapshots and/or slow metadata IO, we may be able to implement some caches, albeit that's naturally quite finicky, especially if we do not want to lose all that info we can now show, and so it's not just a small patch.

FYI, there's a bug report to track this https://bugzilla.proxmox.com/show_bug.cgi?id=3752

 

[dekiesel]

Restoring via pct restore... is not impacted by this, right?

I am using remote storage to secure my backups so atm I can't restore from the GUI atm.

 

[t.lamprecht]

Restoring via pct restore...is not impacted by this, right?

No, there you already pass an explicit backup, so it doesn't need to list all backup groups with all their backup snapshots, meaning it has much less work to do.

FWIW, this issue often comes from older setups that didn't have namespaces inside datastores available when they got configured.
Nowadays, one can create a namespace hierarchy and doesn't need to have all of their (possible huge amount) of backups at a single level. This can help to reduce the read-IO from the PBS datastore, as we can relatively simple list only the backups that belong to a specific namespace.

 

[dekiesel]

That's good to know, thanks!

Atm i am only using two namespaces and dev and prod. So I'll create a ns per vm/ct