Maxing out one pool impacts others

[marmoset]

Hi,

I have a proxmox install with multiple pools. The ones in question are:

bulk (8x6TB spinning SAS raidz2)
nvme (2x2TB nvme mirror)

I am copying in some images from a remote host:

dd if=/dev/whatever bs=1M |ssh root@proxmox dd of=/dev/zvol/bulk/<ID> bs=1M

and as that bulk pool gets busy, all the VMs (that only use the nvme pool) are non-responsive.

It feels like there's some shared resource that crosses pool boundaries, does this sound familiar to anyone?

 

[itNGO]

Hi,
have a look at CPU-Load on the Node. Possible that IOWAIT and/or CPU-Usage is unusual high.
Maybe it is possible to limit throughput of the copy. Spinning disks on raidz2 without SSD can slow down the whole system when under load.
Also check memory usage. ZFS by default will take up to 50% of memory.