Ritardo IO

[toto-ets]

I have an IO delay problem, I have 2 local datastores with raidZ, 1 VM on datastore 1 with win10, and a VM on datastore 2, when I use the VM on datastore 1 and download files the IO delay increases up to 70%, but at the same time the VM on datastore 2 also slows down, why does this one also slow down if it is on another separate datastore?

 

[UdoB]

Please tell us something about the storage devices (rotating rust? NVMe only?) and show us your datastores cat /etc/pve/storage.cfg and the topology of the ZFS pools zpool status.

Basically: if you use a shared resource by two processes then one process may affect the other. In our context this can happen if you have only one set of HDDs instantiating one single pool.

Yes, you can have several Datastores on one ZFS pool. On purpose or accidentally - with "local-zfs" using ZVOLs AND "local" as a "Directory"-store for example.

 

[toto-ets]

here is my conf. I have 2 separate pools with 2 raidZ, with rotating disks

dir: local
path /var/lib/vz
content iso,vztmpl,backup

pbs: PBS
datastore Datastore
server 192.168.21.150
content backup
fingerprint xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
prune-backups keep-all=1
username root@pam

zfspool: Datastore1
pool Datastore1
content images,rootdir
mountpoint /Datastore1
nodes pve1,pve3,pve2

zfspool: Datastore2
pool Datastore2
content rootdir,images
mountpoint /Datastore2
nodes pve3,pve1,pve2