10-11% I/O delay

[FuriousRage]

I see on my node that it has IO delay of around 10%, is this normal?
Root is mirrored nvme. all the VM:S are raid10 7200rpm mechanical drives.


--

Also the "status" part (top) keeps having the "loading" spinner,

 

[Dunuin]

Yes that's normal. HDDs are terribly slow. Like 1/500 to 1/25000 of the IOPS performance a good NVMe SSD could handle. Once you hit HDDs with random reads/writes (what you usually will do when using them as a VM storage) the IO delay will go up and your system performance down.

 

[FuriousRage]

Yes that's normal. HDDs are terribly slow. Like 1/500 to 1/25000 of the IOPS performance a good NVMe SSD could handle. Once you hit HDDs with random reads/writes (what you usually will do when using them as a VM storage) the IO delay will go up and your system performance down.

Hi.
The Debian Guest with NextCloud, when i start to sync (i have like 8TB to sync) that guest becomes quire unresponsive and load average around 8.x
Guest have 12 GB ram, 4x4 cores/threads.
Guest resides on the mechanical raid10 (4 drives)

My precious server i rented at hetzner, have 2 mirrored mechanicals, also NC guest, and that one syncs alot faster.

 

[IsThisThingOn]

Your guest sync 8TB of data from NC?
So your guest are using a 8TB VM disk?

I don't know your use case, but is there a reason why such a huge amount of data has to be on a zvol instead of a dataset?

Does your Hetzner server and your new server have the same HDDs and Raid controller?

 

[FuriousRage]

Your guest sync 8TB of data from NC?
So your guest are using a 8TB VM disk?

I don't know your use case, but is there a reason why such a huge amount of data has to be on a zvol instead of a dataset?

Does your Hetzner server and your new server have the same HDDs and Raid controller?

After hours and hours of googling and testing.
What im 98% sure fixed, was to change the hardware to virtio and the disks to "Default (no cache)", now its speedy speedy.