Hello!

 

TLDR version;

After upgrading from Unraid 6.1.9 to 6.3.5 I have had several complete lockups of the unraid tower. I believe it is related to I/O to the array, as it tends to happen while I host 7Days2Die-server within my Win10 VM, while dockers run directly on my Unraid tower. Never had these problems prior to upgrading to Unraid 6.3.5

I still run the stock BIOS on my mainboard, but are strongly considering flashing it to a more recent release (Gigabyte GA-Z170XP-SLI).

Would like some input on how to approach this.

 

Long, detailed version:
I build my one and only UnRaid-server about two years back. It´s basically a 10TB array (3 WD Red drives; 3+3+4TB) with one parity (also WD RED 4TB) and a 128GB SSD (HyperX Fury) cache.
The hardware is:

Gigabyte GA-Z170XP-SLI motherboard

Corsair TX 750W PSU

Intel Core I5-6600 Skylake CPU

Crucial DDR4 2133MHz 32GB (4x8GB)

Creative Soundblaster Audigy FX soundcard (passthrough to Win10 VM)

A secondary PCI Intel Pro/1000GT NIC

One monitor plugged directly into the motherboard VGA-output for Unraid console

Another monitor connected to the PCI-express GPU for passthrough to Win10 VM.

A keyboard for console.

A keyboard and a mouse passed through to the Win10 VM.

 

I´ve kept the amount of "server roles" on a minimum, only running a few dockers, the main one being PlexMediaServer. On top of this I´ve built a Win10VM with GPU-passthrough for work and gaming. This have all been running very very stable, and I cannot recall any scenario in which i had to force a restart on the original build running Unraid 6.1.9.

 

Then, a couple of weeks back problems started: 

* A friend donated a used 'EVGA GeForce 1060 3GB', which I decided would replace my old trusty 'Ati Radeon HD5850'. One friday night I decided to start working on this, I decided to take the big leap from UnRaid 6.1.9 to 6.3.5 through the 'upgrade'-option under the Plugins-tab. At first reboot after this, I had no network connection to the tower. However another reboot seemed to fix that issue.

* I then switched the GPUs, configured my VM accordingly and things seemed to work as it should.

* The following days I tried building another VM, this time with osx with VNC-connection. (Although still not working, this is something I'm hellbent on accomplishing.)

* Around this point I experienced my first real tower crash while working inside the Win10 VM desktop. Network connection was cut, and so I didn't have any other option than to push the reset button my tower - something I really hate doing. At this point I read that VMs should be configured with OVMF bios, not SeaBios. This should secure that my console monitor/keyboard/mouse would be operational even after the VM were booted, and should give me the possibility to read errors and safely reboot the tower if need be.

* Well, last night I had a new crash just after a scheduled file backup from inside the VM to a mapped network drive on the array was finished. Complete lockup both inside the VM and on the console. As the crash occured, i noticed the console monitor flashing all sorts of text. When it stopped typing, this is the image i got from it before once again being forced to press reset:

 

It seems lockups occur when the tower is doing some rather heavy read-write operations to the array. I tend to host a 7Days2Die-server inside the Win10 VM for my wife and a friend, and have always thought that hickups or crashes inside the VM would not affect the Unraid host itself. Last night I thought through this again, and then it occured to me that since the 7D2D-server is stored on the array and mapped inside my VM, the harddisk management or the intensity of read writes to the array could potentially crash the tower even if it was initiated within the VM itself.
 

All in all I feel that the upgrade from 6.1.9 to 6.3.5 is the cause of the problems, or at least have given the problem means to have an devastating affect on my rig. I've tried googling for similar scenarios, but I have not found any cases which match my experiences or are close enough to my build. Last night I found a tip regarding Marvell sata controllers being a source of many headaches, but I haven't found any information suggesting that my mainboard have Marvell sata controller ports.

 

One theory I do have - which I feel I just need an 'approving nod' from the experienced users on these forums on before pursuing - is that the introduction of Unraid 6.3.5 somehow touched upon a bit dodgy code in my BIOS and causes the system to lock up due to some IO-glitch of some sort. I am strongly considering upgrading the BIOS, as I am still running the BIOS that the motherboard came with (released in 2015). Either that, or I just need to downgrade back to Unraid 6.1.9. We keep the photos and homevideos of our children on the PlexMediaServer, and I need to trust the Unraid + online cloud backup keeping my data safe.

 

Please, someone advice me on what to do. I have a dump of the Diagnostics (taken after the reboot) if anyone would like to have a look at them.

 

For the record; I have done memtest. After 3 successfull passes i cancelled the test.

Also, how important is it to use the most recent 'VirtIO-drivers' inside the VMs, and could old/faulty drivers cause complete crash of the Unraid host?

 

(Edited for additional information/corrections.)

Edited December 14, 20178 yr by blodfjert
additional information/corrections. TLDR-version added.

9 minutes ago, blodfjert said:

We keep the photos and homevideos of our children on the PlexMediaServer. I cannot afford losing that kind of data.

You are keeping one or more backups of important data, with at least one copy at a different location?

 

A RAID - any kind of RAID - isn't a replacement for a backup. It's just a way to improve the availability in case of hardware errors - or in some situations a way to increase the total capacity/performance.

3 minutes ago, pwm said:

You are keeping one or more backups of important data, with at least one copy at a different location?

 

iCloud keeps me safe for now, and during the month I'll switch my 12/2MB-connection for 100MB syncronous fiber connection, at which point I intend to subscribe to some cloud backup solution for my entire archive.

 

 

Today I upgraded BIOS to version F22C, released only 14 days ago.

No hickups yet. Fingers crossed.