Hi All,aber das backup wird auf PBS seite fertig angelegt (nur ohne client log)? kannst du mal bei einem backup mittels "atop" auf beiden seiten "zuschauen", ob beim finalisieren des backups die last nach oben schnellt? eventuell macht es sinn das backup haendisch in der shell anzustossen und in 'ts' (paket: moreutils) zu pipen (alsovzdump .... | ts
), dann ist der output gleich mit timestamps versehen zur leichteren korrelierung.
After having had quite some trouble with PBS I decided to do a little test...
- Virtual PBS Server
- iSCSI Volume for PBS Data (on a dedicated backup storage that is replicated)
- All attached over 10G
Turns out the problem seems to be with my physical Server System that was playing havoc with the internal SATA Drives.
When running multiple backups of larger VMs, the physical server developed system-load figures of 25+ - which is absolutely not ideal.
With the virtual machine (I backup using pools now so I can give the PBS a dedicated Pool to be backed up using vzdump) I have not had any problems so far. Backup Speeds have gone up to 5GBit/s+ (depending on the VM and the storage the VM lies on).
This is now running for several weeks - and hasn't missed a beat so far.
So... the backup server absolutely needs IOPS - and a system being able to provide IOPS without generating high loads.
I think I will keep it as it is now...
Thanks!
Tobias