Because you have a very high io time on average. My question:
Do you also have the problem that your server is simply restarted after more than 9 days? Because I currently have the problem with an RS 8000 with a 3 TB disk with Debian 11 (bullseye) and Rocky Linux 8.x.
The I/O time is high because the server is capping the write speed and hence delaying the writes.
And no, I have not had the server up 9 days yet. I sincerely hope it will not reboot, that's definitely not by design.
If that is so, the selected settings are not suitable at least for the bigger RSes. Or it is really a technical issue, as it has not been like that before.
I understand if there is some limit set (since the disk pool is shared between RSes), but it does feel very low.
Best guess: NetCup configures the VMs using something like
Codethrottling.iops-total=... # equivalent of 40MB/sek throttling.iops-total-max=... # burst of 200MB/sek throttling.iops-total-max-length=... # 20min
Reference: https://github.com/qemu/qemu/blob/master/docs/throttle.txt
That is my guess too. I just wanted to confirm if that is the case.
Check the disc hypervisor, should be VIRTIO but is SCSI on most servers.
Makes a huge difference in performance, could come with some risk.I've used it for some years by now.
I run one on VIRTIO and one on SCSI driver. Same results on both machines.
Hi guys!
I noticed that my server (RS 8000 G9.5) sometimes gets capped iops to 40 MB/s. Is that expected? Normally it can go a lot higher than that, but sometimes that's a hard ceiling. Is this a limit Netcup is applying?
Screenshot 2022-11-08 at 13.30.17.png
At the same time the server gets capped, the IOWAIT goes up substantially (which is expected).
Screenshot 2022-11-08 at 13.31.05.png
Can this be affected in any way, or do I need to find another host that doesn't do this?
