TrueNAS 'truly' stable on Ryzen?

[traCk]

TRUENAS SCALE

Trying to upgrade/replace my Haswell Xeon CORE system has become a huge headache

Ryzen looked like a great proposition. I have a few chips laying around
Went full in, my goal wasn't to cheap out
AsRock Rack X470D4U, 16GBx2 Samsung ECC strait off the QVL
Starting with a Ryzen 2700, system would hang randomly. Especially when dealing with Docker 'Apps' or VM's

Then I start searching, nightmare threads of LOTS of people having issues with Linux/BSD. Lots of 'partial' fixes by turning off C-States, turning off Global C-state control entirely, Changing PSU Idle power settings. And in more these threads it seems people 'gained' stability, but didn't fully resolve their issues

Wish I would have seen these going in

Looked like most issues were attributed to Ryzen 1XXX/2XXX....
Ok, pop a 3700X in the system, popped in the latest bios (Which was previously not installed due to Ryzen 2XXX incompatibility)

The system will randomly hang, especially with Linux VM's. Only difference is the system automatically reboots. On the Ryzen 2700 the system would hang and have to be manually power cycled.

The more I look the more I see people claiming they can't fully resolve lockups

Is Ryzen just not ready for a non-windows system management yet?

 

[anodos]

We'll have a nightly out pretty soon with kernel 5.15 (which has better support for ryzen). It might be worth trying.

 

[traCk]

We'll have a nightly out pretty soon with kernel 5.15 (which has better support for ryzen). It might be worth trying.

That's awesome. Looking at 5.15 improvements I do see plenty of AMD fixes, nothing specifically for hangs or crashes. Is this something you see as part of the fix? Trying not to get my hopes too high if the fixes don't really relate to those issues. I do see power management improvement, which definitely seems like its part of the issue

 

[anodos]

I haven't looked into anything in particular, but generally newer kernel versions have better support for newer hardware.

 

[LarsR]

I've used a 1600x for about a year on core and switched to scale and a 3700x on an msi x370 board.
Yes its gaming hardware but it works for me and i got it laying around after i updated my gaming pc.
1. and 2. Gen ryzen on core needed some bios tweeks like disabling global c-states, erp-ready and amd cool&quit.
After switching to the 3700x and scale i enabled them just to see if 3. gen still has the same problems, but to my surprise
i don't have any issues.

 

[traCk]

I've used a 1600x for about a year on core and switched to scale and a 3700x on an msi x370 board.
Yes its gaming hardware but it works for me and i got it laying around after i updated my gaming pc.
1. and 2. Gen ryzen on core needed some bios tweeks like disabling global c-states, erp-ready and amd cool&quit.
After switching to the 3700x and scale i enabled them just to see if 3. gen still has the same problems, but to my surprise
i don't have any issues.

I wish that was the case for me. The system will stay up days at a time, but if you are spinning up/down VM's or upgrading docker apps is when it likes to show its true colors. 'Static' stability seems fine, but I don't trust it to not have that 1 crash that corrupts my pool.

 

[LarsR]

I've bin on scale since the late alpha, i'm using 16 apps from the truecharts catalogue and had a manjaro and win10 vm running for some time and had no issues at all. I deleted the two vm's cause i no longer need them and it was just to play around with anyways, but i'm still using all 16 apps at a time.
For testing sake i could spinn a vm back up and let it run for some time just to see if my system will have any problems.