Unraid 6.9.2 crashing overnight

[unraid-raider]

I've been using this server pretty successfully for a few months now.

 

I made one recent tweak which was to set all of the shares as allocation method: most free, but I can't imagine why this would have an impact on stability?

 

It's happened twice now that I come to the machine, there's nothing on the display, but the fans are humming and the lights are on, it doesn't respond to ctrl+alt+del nor a quick press of the power button, only option is to hard power off and then back on again.

backup-diagnostics-20210801-2207.zip

[trurl]

26 minutes ago, unraid-raider said:

allocation method: most free

Probably the worst for performance but shouldn't crash.

 

https://wiki.unraid.net/Manual/Troubleshooting#Persistent_Logs_.28Syslog_server.29

 

Also

 

[unraid-raider]

Thanks, I've setup syslog, so will wait and see what I capture next time.

 

I'm curious on your performance comment, I'm aware of the potential around overlapping parity writes, for me this is less important than unraid filling a single drive in my multi drive array and then if that fails the whole full drive needs to be re-created. I'd rather a failure state was an equal load on the system rather than russian roulette of it hitting the one or two full disks. Is there anything else you're thinking about?

 

I had read the ryzen part of the faq, however given this has literally been running flawlessly for almost a year now (11 ish months) so I figured it was unlikely to be related to that. Perhaps the syslog will shed some light on it.

 

No crash last night, but it doesn't happen like clockwork so will keep an eye over the next few days.

[itimpi]

1 hour ago, unraid-raider said:

I'm curious on your performance comment, I'm aware of the potential around overlapping parity writes, for me this is less important than unraid filling a single drive in my multi drive array and then if that fails the whole full drive needs to be re-created. I'd rather a failure state was an equal load on the system rather than russian roulette of it hitting the one or two full disks. Is there anything else you're thinking about?

 

Just z few points of clarification  

• If you are using the default High Water allocation method and reasonable values for Minimum Free Space and Split Level in the share settings then there is no reason you should be running into disk full issues on one drive with other drives relatively empty.
• If a drive fails it will take the same amount of time to rebuild that drive regardless of whether it is 1% or 99% full.