Best way to downgrade to Dragonfish from Electric Eel - and questions

TrueNAS General
1

Hello All

Things have gone wrong and I would like to find out (a) how I ought to “downgrade” and (b) why things went wrong and (c) how to best manage “apps”.

(a) I read up on the state of play regarding EE (Electric Eel) and, when I was confident I could do it with minimal impact, I upgraded yesterday. The u/g worked flawlessly in terms of the “OS” and all my data is intact (and it was already snapshotted and backed up) and the new OS is running fine.

However - I see my CPU hitting 99% very frequently which seems to be due to a couple of the apps which I am running (on BlueFin and then Cobia and Dragonfish and now on EE). One of the apps is frigate and another is motioneye, both ingesting and then serving out CCTV video streams. Previously the general CPU load was below 50% when runing these (and other) apps. This increase is causing me concern and is the reason why I think I should downgrade.

I tried downloading the update file and running it but that didn’t work so

Question 1 - is a bootable USB stick the way to go and simply do a fresh install (I have lots of configs backed up )?

But …

(b) … the TrueNAS upgrade process also caused (or anyway required) the frigate app to upgrade too, so that the image it uses is a more recent one so I am now not comparing like with like (maybe the earlier version of frigate would impose a lower CPU burden on EE the same as it presumably did on Dragonfish). Should I have deleted my apps and reinstalled them, instead of relying on the ugrade process to migrate (I know that that should have worked OK but it seems not to have done so).

Also - prior to and continuing after the upgrade I have used using the Stux method of running Dockge which appears at an IP address one below TrueNAS and was a reliable and easy to understand method, moving me away from TN’s apps approach. Now that the Apps are Docker based, I thought I culd install dockge as an app and run things inside it, but this seemed to lead to higher CPU use as well.

Question 2: how should I have handled the apps in order to not hit these problems?

And point (c) which is most easily addressed via a third “set” of questions, namely:

3a should I stick with Stux’ dockge approach to run my apps?
3b should I install dockge as a trueNAS app and install my apps within it?
3c should I install frigate directly as an app, and motioneye as a custom app?

To reiterate, the apps which seem to burden my CPU are frigate and motioneye (seen by SSH into trueNAS, running htop, seeing the top loads and then stopping the associated apps and watching the cpu load drop significantly). The other apps I want to run are jellyfin, tailscale, Adguard home and uptime kuma.

Finally - the new UI is great and I will look forward to revisiting EE when I have settled my nerves and got my act together, but I note that 24h after upgrading, this part of the dashboard is still unpoulated:

image
image1109×415 15.9 KB

is it because I need a scrub task and a disk scan to occur? Theyre both scheduled but I can’t recall when for.

2

You can try booting into the previous version in the boot-pool

I think this is the correct section of the documents
https://www.truenas.com/docs/scale/24.10/scaletutorials/systemsettings/managebootenvironscale/#boot-environments

1 Like
3

That’s clever! I didn’t think of that approach. I’ll look into it and I’ll report back. Thanks for this idea.

edit, a day later: @SmallBarky 's solution worked. Excellent! Thanks for pointing it out (and thanks to TN for designing it in).

Rather than work through all of the seven hundred and forty eight combinations to find out what the problem was, I tried a dozen approaches and then gave up, simply reverting to Dragonfish instead which has given me the apps I need at a burden of a nominal 30-40% CPU instead of the 90+ I was suffering with EE.

1 Like
4

From looking around the web a bit, it seems like for both Motioneye and Frigate misconfigurations can lead to high cpu usage. It might be worth your time to compare settings (on TN and in app) between the two deployments and see if there are any differences. If you were able to roll back to the DF boot environment, I’d expect you should be able to do so to the EE one as well. You might be able to reach out for help upstream at the app projects themselves.

It’s possible that there is a bug related to how we’ve migrated the settings or set defaults on the new catalog. It’s also possible that they’re configured the same but one of those settings responds differently on docker than it does on the K8s chart.

Edit: Based on your other thread, the switch in Frigate versions may have contributed to this as well. The EE catalog is on 0.14.1, so if you’re comparing that to a 0.13.X version in DF that may play in. Based on a quick search there were a few reports of increased CPU use after 0.14–though it seems to be very system and settings dependent.

5

@DjP-iX this is very interesting and encouraging to read. I am going to take a break from any significant fiddling around with things for now but you have some good advice for me to follow when I start trying to get to the bottom of things. I’m submitting a bug report as best I can, after the event.

1 Like