Odd behaviours with Unraid CPU Usage

[Xyler94]

Hi There,

After upgrading to version 6.12.4, I've noticed my unraid's WebUI being slower than usual, and the CPU Spikes to 100% more often than usual.

However, when looking at the HTOP output in the CLI, it doesn't match the CPU usage of the WebUI, so I'm sort of confused as to the issue. I know the WebUI and HTOP outputs don't refresh together, but the attached screenshot, the WebUI was like this for a few seconds, and the HTOP output refreshed several times before I took the screenshot.

None of my Dockers or VMs are being sluggish, and none of them are showing signs that they're taking up system resources like this. 

I've attached a screenshot of the WebUI and HTOP views together.

Was curious if this is normal behaviour. If it isn't, what are steps I can take to troubleshoot this, because I'm running low on options on things I've tried. Even rebooting doesn't seem to fully help.

It isn't always at high utilization, but it does so often enough, and the WebUI is sluggish due to it. 

 

 

[JorgeB]

Dashboard also includes iowait, htop doesn't.

[Xyler94]

14 minutes ago, JorgeB said:

Dashboard also includes iowait, htop doesn't.

That would tell me the differences between the two, but that doesn't tell me why unraid seems less responsive than usual.

I've been trying to figure why my unraid is spiking often. If you're saying the difference between these two may be iowait, then do I have an IO issue somewhere?

[JorgeB]

iowait will usually show when there's something writing to the array or a pool, diags may show more info.

[Xyler94]

29 minutes ago, JorgeB said:

iowait will usually show when there's something writing to the array or a pool, diags may show more info.

I downloaded the logs (I attached it here if you're curious).

What's odd to me that stuck out was in the "SMART" logs. That top "NVMe" doesn't look right, I only have 1 NVMe drive right now in my system

 

For some reason, it shows the Sabrent Rocket NVMe drive twice in this folder, with different names. I checked both, and they're the same serial number.

If this is simply my PC slowing down, I'm gonna be upgrading my server in the coming weeks, but I am curious what's causing this. 

xylerserver-diagnostics-20231115-1055.zip

[JorgeB]

1 hour ago, Xyler94 said:

What's odd to me that stuck out was in the "SMART" logs. That top "NVMe" doesn't look right, I only have 1 NVMe drive right now in my system

You can ignore that, it's normal, do you know what's writing to the array? In the diags there's something writing to disks 2 and 3.

[Xyler94]

12 minutes ago, JorgeB said:

You can ignore that, it's normal, do you know what's writing to the array? In the diags there's something writing to disks 2 and 3.

Okay, as long as it's normal, I'm not worried then.

Disk 2 seems to contain my Home Assistant VM.
Disk 3 contains my NextCloud VM (Although I would have liked that on Disk 1, I'm slowly upgrading my disks to Seagate Ironwolf drives)

Disk 2 would make sense, Home Assistant logs many things, and there's a power monitoring device in the Zwave network I got going, and it constantly reports power usage.
Disk 3, NextCloud would be reading and writing to synchronize with my desktop and phone for cloud storage.

I'm now curious if it's NextCloud then. Or if someone's trying to bruteforce onto NextCloud.

Other than that, I'm not writing anything to my array at the moment. 

[JorgeB]

Having VMs on a HDD array with parity will not be optimal for performance, and very likely the reason for the iowait, see if you can moved them to a pool

[Xyler94]

Just now, JorgeB said:

Having VMs on a HDD array with parity will not be optimal for performance, and very likely the reason for the iowait, see if you can moved them to a pool

Like a Cache Pool?

I did notice my Windows Server VM being odd when it was on the array.
However, some stuff doesn't need the pure speed of the SSDs, and needs big storage. NextCloud for instance needs big storage, not necessarily big performance.

I should move some things over to the SSD Cache Pool though. I should try that. There are a few VMs I could put on the Cache instead.

Question though, should it be Cache only or should I do a Cache to Array with mover?
I had mover enabled on my Cache domain once, and it caused oddities with Windows Server service game services, kept crashing the game server (minecraft especially).

[JorgeB]

Cache only, and to move the VMs you need to stop them and disable the VM service first, libvirt (and docker image if you use it) should also be moved to a pool.

[Xyler94]

1 hour ago, JorgeB said:

Cache only, and to move the VMs you need to stop them and disable the VM service first, libvirt (and docker image if you use it) should also be moved to a pool.

Hi there,

After moving the VMs I could to the cache, and doing other things like you suggested, it seems to have gone okay, but now one of my VMs aren't booting up properly.

It simply boots to this, and now I'm worried I lost the data after moving it.

I cut and paste instead of copy paste the files. While it's not technically a huge loss if I lost this VM, it was my Home Assistant VM, and setting that back up would be a nightmare...

Thanks for your help so far though, I do really appreciate it!

[Xyler94]

Never mind, I just redid the base VM and it came up no problem.