Let me describe a setup I prepare for a customer.
So far I see less performance than expected or wanted, so I'd like to
check for things I miss.
A customer runs one single Proxmox Virtualization host (PVE).
YES, I know that I could also ask things in a Proxmox Forum ...
That host has plenty of CPU and RAM and a large (for me) array of SSDs.
I don't know the actual RAID level right now, but it should be "fast
enough".
So far there are maybe 10 VMs, not much.
One is a PBS (Proxmox Backup server), it should write VM backups to an
LTO7 library attached.
As I noticed that I can only attach the SAS-card to a single VM I
decided and managed to install Bacula into this PBS-VM as well. Works.
(no, no conflicting jobs right now. pbs does nothing with tape so far,
disabled that while setting up)
A second VM will be a fileserver. Debian 13.4, bacula-fd inside.
Data from an external legacy server was rsynced in, now I have set up
the fd, plus some filesets and jobs.
Things work, but I only see around 50-70 MB/s for backups.
I test with and without Data Spooling, doesn't make much difference.
Both VMs are set up with virtual NICs on the same virtual bridge ... so
in theory they should have 10 Gbit/s network (virtio nics in place, yes).
The ressources shared are storage (one VM reading from the array, the
other writing it to the same storage, just another virtual disk = spool
area), CPU and RAM, right?
Nothing looks topped out or so ..
When I disable Data Spooling for the Job, the storage conflict should be
gone, right?
I know that this setup isn't exactly prepared for top performance, but
anyway I'd expect more throughput possible, right?
One job backs up 3.2 TB alone .. this takes days with the current speed.
Let me know your thoughts and what parts of the config I should look at,
or maybe share here.
Thanks!
_______________________________________________
Bacula-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/bacula-users
