yesterday I choose to upgrade my TrueNAS Scale system from the latest Cobia to Dragonfish Scale 24.04.2 version. I read the pros/cons about the Truecharts direction - but I was able to finally move everything to TrueNAS repo apps without any issues. Only one thing what I found that after the upgrade, my CPU utilization and power consumption is increased (i5-13500k, 32GB RAM, 3x8TB disks). With CPU increase, CPU temperature is also high, which causing a noisy fan of course.
With more apps in overall, my former Cobia setup was around 5-7 CPU%, now it’s increased to 15-17%. Normally, I ran with auto-update, so I didn’t changed anything on system level, like ZFS, cache, or else.
I read various topics about the new kernel, ZFS Cache/swap issues but all of them were related to Dragonfish 24.04.1 or beta. Is still happening with 24.04.02 ? I’m seeing a ~10% resourse increase with the same (or less without traefik, etc.) apps comparing to Cobia.
Could you give me some directions what I should to optimize, change, or applying a fix for it ? Thanks !
After auto-update the system restarted itself - I did not restart my NAS after that.
I really no have currently any utilization-heavy app. Cloudflared, plex, ddns-updater, nextcloud, pihole and shinobi which are currently active. I really didn’t would like to go with other apps until i’m not fixing this somehow.
Before the peak it was the Cobia, the peaks shows the update and the app installs after and clearly visible, that now I have a higher CPU utilization overall with less apps than before.
ZFS Arc size seems to be fine, there’s no any clue that this is problematic. However, at Memory utilization:
There are some differences between Cobia and Dragonfish:
Swap: Cobia: enabled by default, Dragonfish: Disabled by default (because ARC is effectively the memory manager) ARC: Cobia: max 1/2 memory, Dragonfish: no maximum
In terms of what you are seeing:
CPU graph: I have no explanation. I would use the shell and sudo -i and then run kubernetes or Debian commands to look at which processes / containers are using the most CPU.
Memory usage: I am guessing that the change of apps has resulted in less “Used” memory, and this is now taken up and used for caching.
I’ve just found what caused this cpu increase. I remember, that I shutted down my VM and disabled the autostart option before I prepared the upgrade to Dragonfish.
When I investigated the processes from shell using the command: “top”, just saw that qemu eating too much cpu time - then I checked my VM state and just saw that somehow this particular vm is running and consuming CPU, since TrueNAS start. I did a quick shutdown and everything dropped back to normal - before I wanted to change anything. Strange.
