UNRAID 6.6.6 - Soft Lockup - Ryzen 1800X - Asus X370 Pro

[itimpi]

1 minute ago, Marshalleq said:

Seems the logs are wiped at every reboot.  Surely not, but that's what they're telling me.  

Yes.   UnRAID is runs purely from RAM and this includes the log files.    Unless special action is taken to write them to persistent storage they are lost on a reboot.

[Marshalleq]

Thanks for the (lonely) reply in this thread - I think I'm beginning to talk to myself....

 

OK, so I will pull the USB like was suggested in another post which actually doesn't really sound like they're in memory though....  Syslog might get close to covering it if I could enable that (think I saw that somewhere) but I don't trust it for a crash really - will probably miss the important last entires due to network being down I'd guess.  Any advice?  In the mean time I'll do a search to see if I can put them on a drive.

[Vr2Io]

I have this board with Unraid and no unstable issue ( but this board have some issue to auto config memory speed/timing especially with 4 sticks ).

You have 32GB memory, what is memory spec. and actual running speed/timing.

Pls in command prompt provide "dmidecode -t 17" info.

[Marshalleq]

Hello, thanks for your input, I have thought about RAM a few times, but haven't yet been able to fault it as far as I know.  I have enabled the USB logging for the interim and have attached output of dmidecode thanks.  The RAM is https://www.gskill.com/en/product/f4-3200c14d-16gfx

The RAM sticks were purchased at two different times with the second lot bought just because I started running UNRAID and figured I could use it.  It's MEANT to be the same and as far as I can tell it is, but could be wrong.  Thanks again for your help - this is driving me mad.

dmidecode.txt

[John_M]

Have you run MemTest for 24 hours or so?

[Marshalleq]

I have on the first set and a lesser amount of the second set - but I guess I could try it again with the whole lot.  I've added more drives and things - it's a great power supply, but maybe it will all amount to something.  Do you know, do I have to run it for a whole 24 hours or can I run it for 12, watch some files, then run it for 12 again?  This is the downside of migrating from my QNAP, a lot more effort and downtime when I have to mess with things...

[Vr2Io]

You have 4 3200Mhz stick and all should be single rank, if speed over 2600Mhz then will run in 1.35v, I suppose it already correct auto config that, pls check (1).

BTW, 4 sticks at 3200Mhz may have problem, suggest clock it at ~2600Mhz and try. ( Just change it at BIOS )

[Marshalleq]

OK good idea.  I'll check the voltage and increase if it's not correct it.  If it is, I'll try dropping the speed to standard rating (not DOCP).  Why do you say 4sticks at 3200MHz may have a problem?  Is it a common thing?

[Vr2Io]

1 hour ago, Marshalleq said:

OK good idea.  I'll check the voltage and increase if it's not correct it.  If it is, I'll try dropping the speed to standard rating (not DOCP).  Why do you say 4sticks at 3200MHz may have a problem?  Is it a common thing?

Because I have 48GB in 4 sticks ( all stock 2400 ), 2 16GB (double rank) and 2 8GB (single rank), different BIOS will have different behaviour, my current BIOS ver will wrongly clock them at 2600 2666 (1.2v) and system would crash, test show it only stable at 2400, but I actual run them at 1866.

 

* remark : if only 2 single rank stick, I can run it stable at 2600 2666 *

Edited February 9, 2019 by Benson

[Marshalleq]

It was on 1.35 so I've dropped it to 2666Mhz.  It doesn't have straight 2600.  Also I left it on DOCP and just changed that part.  Hopefully that's good?  I don't know much about the memory side obviously!

[Marshalleq]

I assume I could also up the memory voltage.  But anyway, will try it at slower speed for a while just to see if it makes a difference.

[Vr2Io]

1 minute ago, Marshalleq said:

It was on 1.35 so I've dropped it to 2666Mhz.  It doesn't have straight 2600.  Also I left it on DOCP and just changed that part.  Hopefully that's good?  I don't know much about the memory side obviously!

Pls stay and try, good luck.

[Vr2Io]

1 minute ago, Marshalleq said:

I assume I could also up the memory voltage.  But anyway, will try it at slower speed for a while just to see if it makes a difference.

1.35 alraedy got RAM warm, suggest don't take higher, DDR4 memory seems easy die, I have meet several memory die case.

[Marshalleq]

OK thanks.  Makes me think I should have got cheaper RAM, but more like 64GB.  Oh well.

[Vr2Io]

59 minutes ago, Marshalleq said:

Do you know, do I have to run it for a whole 24 hours or can I run it for 12, watch some files, then run it for 12 again?

When I build new system, I like test it under Windows by different stress tools, usually less then 6 hrs and all show fine. So I won't got trouble relate any stable issue.

 

59 minutes ago, Marshalleq said:

This is the downside of migrating from my QNAP

I am a QNAP friend too, but due to drive bay need and C/P, I turn my storage by Unraid ~3yrs ago. ( Also turn a ATOM Qnap to Unraid too, but now not use it anymore due to slow CPU and platform I/O )

 

37 minutes ago, Marshalleq said:

Makes me think I should have got cheaper RAM

Agree, but I also buy 4 8GB 3000Mhz for future build, may be Threadripper or Intel.

 

48 minutes ago, Benson said:

Is it a common thing?

Yes for Ryzen

Edited February 9, 2019 by Benson

[JorgeB]

With Ryzen you should keep to the max non overclocked RAM speeds depending on your config, or risk instability and sync errors:

 

[John_M]

It's all about the number of physical DRAM chips that are hanging across the bus because consumer motherboards use unbuffered DIMMs. The more there are, the more difficult it is to drive them at high clock speeds. Johnnie's table is correct for first generation Ryzen, such as your 1800X. For second gen. everything moves up a step, so the left hand column would read 2933, 2667, 2400, 2133. For third gen. it's rumoured that everything moves up another step but as the speed of the Infinity Fabric will be decoupled from the RAM clock, slower memory will supposedly have less of an impact on overall performance. Presumably Intel processors have their own limitations when it comes to driving heavily loaded memory buses at high speed though overall performance is less dependent on memory speed.

[JorgeB]

16 minutes ago, John_M said:

For second gen. everything moves up a step, so the left hand column would read 2933, 2667, 2400, 2133.

Not always, depends on the board's number of DIMM slots and PCB layers:

 

 

[o0atomos0o]

Curious if this Marshalleq ever got this issue resolved? I'm currently running a new build with AMD 1800x and Gigabyte AX370 Gaming K7 board. I'm trying to run my initial parity-sync and my server locks up (soft lock?). The web GUI will be down and the server won't send signal to the monitor but the discs are spinning.

I've tried disabling c-states, ensuring memory is set to auto, and set power control with no luck. I'm now trying parity-sync with rcu_nocbs=0-15 but not feeling very optimistic.

[trurl]

Probably better to start a new thread with your diagnostics than post to a 5 year old thread about a version you don't have.

[Marshalleq]

2 hours ago, o0atomos0o said:

Curious if this Marshalleq ever got this issue resolved? I'm currently running a new build with AMD 1800x and Gigabyte AX370 Gaming K7 board. I'm trying to run my initial parity-sync and my server locks up (soft lock?). The web GUI will be down and the server won't send signal to the monitor but the discs are spinning.

I've tried disabling c-states, ensuring memory is set to auto, and set power control with no luck. I'm now trying parity-sync with rcu_nocbs=0-15 but not feeling very optimistic.

Sorry I don't really recall.  I know I eventually sold that whole setup - but not because it didn't work - which makes me think it was working fine.  I'd recommend updating the BIOS to latest, making sure the C states etc for the power supply are disabled (there are two settings I don't remember so you'll have to google that), make sure you're running latest unraid, make sure your power supply is good - that is one thing I did eventually change. 

 

Ultimately I ran a Threadripper 1950x for a while which exhibits all the same issues as the 1800x really, then I went to a dual Xeon system in a Lenovo P700 just because the cooling and built in disk options are better.  My sense is the xeon does work a bit more reliably than either of the AMD systems did but that the issues did eventually resolve themselves via kernel updates or something.

 

Another thing I've experienced is an overheating HBA card that can cause a few challenges. 

 

Otherwise, I guess you ask the questions in a new thread as your issue might be slightly different.  I will try to answer your questions there if you @ me.

 

Have a great day!

 

Marshalleq