mjhjmtu

First off, I asked this question 5 days ago over on Serverfault. I hope I'm not doing a bad by bringing it over here to the Unix&Linux Stack. I have also asked this question on 3 other sites not related to Stack, with no answers. I plan on updating each site with an answer, if I can just get it answered. Here we go.

I am having a hard time understanding the output of iostat -x with specific regards to ZFS zvols. I'm running Proxmox 4.4, fully updated and encountering some generally poor IO performance.

While troubleshooting the sluggish performance, I was looking at iostat -x 1 and saw this sort of utilization reading near constantly.

Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s avgrq-sz avgqu-sz await r_await w_await svctm %util
sda 0.00 0.00 77.00 115.00 308.00 640.00 9.88 2.02 10.33 9.92 10.61 3.58 68.80
sdb 0.00 0.00 81.00 116.00 324.00 644.00 9.83 1.32 6.72 6.42 6.93 2.50 49.20
...
sde 0.00 0.00 77.00 117.00 308.00 640.00 9.77 1.16 6.25 5.25 6.91 2.35 45.60
sdf 0.00 0.00 78.00 116.00 312.00 640.00 9.81 1.25 6.45 5.64 7.00 2.47 48.00
...
zd32 0.00 0.00 0.00 197.00 0.00 788.00 8.00 1.09 5.54 0.00 5.54 5.06 99.60

Where I am confused is that the utilization percent for zd32, the zvol of my VM, is at 100%, where the underlying storage is at roughly 50% utilization.

My question is: Shouldn't the zvol utilization reflect the utilization of the underlying storage devices?

For reference, there are other VMs on this system, but this troubleshooting was done after hours, so they were idle. This one VM was the only busy VM, running Windows updates. The zpool is a RAID-Z2 of 7200RPM SATA disks, so not exactly built for incredible speed. I'm just wondering about the utilization right now.

Here are some hints. Yes, it should, because zfs volume is created on zpool which is located on a storage device. If that storage is shared between other resources, they can affect zfs pools and volumes.

Unfortunately, I do not know what Proxmox is, but %util usually shows the time the device has a positive queue of transactions. A number of transactions in the queue is avgqu-sz. Both of these values are also depend on the storage system type and model which can support quite a large queue. So, it may be a bad symptom or not. Therefore first of all it's better to look at: await, r/s, w/s, rkB/s, wkB/s to see if the volume has a real workload and performance issues or not.

There is a special command: zpool iostat to monitor zpool statistic.