High CPU usage on 20 core / 40 thread system after boot.

[gcs8]

So, bit of back story here, last week I had one E5-2630 v4 and 64GB of RAM, boot times were fine, no real bog down after unlocking the encrypted pool. Monday of this week, I picked up 768GB of ram, installed the first 12 sticks (384GB) Monday, and all seemed fine and was for a bit. Tuesday ordered a second E5-2630 v4 for the other 12 sticks and put it all in today after work. After it booted up, all was fine and everything looked good. (Other than FreeNAS does not report two procs in the GUI.) I unlocked my pool and then the server just decided it was going to do its own thing. From ~22:00 to 23:05 it was churning through the drives and keeping high CPU utilisation. After that ~hour it seems to have calmed down. Not sure what caused this strange behaviour, has anyone else seen anything like this before?

Things the box is doing: SMB, AFP, WebDav, Plex Jail, OwnCloud Jail, iohyve (2 VMs each with 2cpu, one CrashPlan one Splunk).

Screenshots of some of the fun
.

 

[gcs8]

Only thing I can think of was a scrub, but zpool status did not show an active scrub. Maybe ZFS wanted to pre-load ARC???

 

[m0nkey_]

My money is that it kicked off a scrub.

 

[gcs8]

My money is that it kicked off a scrub.

 

[Hazar]

Maybe a CrashPlan file scan? It has a setting called "verify selection every x days" and it will generate a fair amount of random I/O while doing that.

 

[gcs8]

Maybe a CrashPlan file scan? It has a setting called "verify selection every x days" and it will generate a fair amount of random I/O while doing that.

Maybe, but I want to say I had not started the iohyve VM for it yet. The VM is also limited to 2 CPUs and is using WebDAV to read the data, which also looks to be single threaded. I have CrashPlan setup to only scan once every 5 days right now, I also have to map the WebDav mount by hand at each reboot on the VM. *shrug*