Select a specific nic to backup/restore

[lxiosjao]

Hi,

It's impossible to ssh into my others VMs during the restore and I dont know why.
I would like to know why all my vms are unreachable when I restore 1 VM.

I use 2 nic :
- moterboard nic (1x eth port : Intel Corporation 82579V)
- pci nic (2x eth : Intel Corporation 82576 Gigabit)
I want to use 1 for backup/restore to/from PBS and the other nic for the VMs.
How am I supposed to do ?
I dont know which nic is really used and when it is.
(I can join the UI on both IP)


Thanks

 

[fabian]

you need to either segment your network properly or add custom routes on both ends (PVE and PBS).. but I doubt it's the network that is causing your VMs to not be reachable, it's much more likely that the restore overwhelms your PVE storage. you can try to monitor with "atop" or a similar tool to see which resources are bottoming out during the restore, and then either apply limits or increase their performance.

 

[lxiosjao]

Thanks for your answer.
It seems to be clear, it's the storage fault.


I made a restore test :
/dev/sdd : the disk where is stored the virtual disk of the vm I'm restoring.
/dev/sda : all my others VMs unreachable

If /dev/sdd is used at 100%, why storage on /dev/sda is impacted ?

(I notice that in atop, sda appears and disappears every 10 sec during the restore, I dont know if it's normal.)

thanks

 

[fabian]

what kind of storage are you using? ZFS can be overloaded in a fashion where writing to one pool can make another pool almost unresponsive for example..

 

[lxiosjao]

what kind of storage are you using? ZFS can be overloaded in a fashion where writing to one pool can make another pool almost unresponsive for example..

ZFS for all my storage.
I see, what can I do to solve this ? (except changing zfs for an other FS )

 

[fabian]

you can rate limit the restore or tune your ZFS setup - but it seems to me (based on the atop numbers) that your system is really underpowered..

 

[lxiosjao]

you can rate limit the restore or tune your ZFS setup - but it seems to me (based on the atop numbers) that your system is really underpowered..

Yes, it's an old desktop PC from 2011.

I would like to change and go to a storage separated from host Proxmox.
My idea is going to :
- 1 cluster (for HA) Proxmox with 3 host (like 3x NUC or something like that with NIC 10Gb)
- 1 dedicated BIG storage 10Gb using ZFS separated (like a DAS/NAS/SAN) for my data (HDD) via NFS

Any recommandation about host and storage ?

thanks

 

[fabian]

well, that really depends on your planned workload (but NUCs are not know for being high-performance, server-grade hardware either ). you can find lots of threads with hardware recommendations and reports on the forum though!

 

[lxiosjao]

Thanks for your advice.
I'm going to search these threads