I upgraded from a cheap SATA board to a full on HBA. Grabbed a Adaptec ASR-71605 on ebay and it keeps crashing my system after a few days of uptime. System runs fine, nothing seems to lead up to to the crash, it's running then it falls over on itself. The adapter is in HBA mode, no BIOS is installed on it, I never got the BBU battery. Could my card be a dud or is it misconfigured? 

I suspect I could be running out of PCI lanes. The CPU is a Ryzen 5800X, and I have a GTX 1650 for hardware transcoding and AI upscaling installed as well. I also suspect it could be unraid itself not liking the card. 

Edited February 5, 20251 yr by tgotr

without diagnostics nobody will be able to make a guess...

Turn on Syslog saving, wait for the next crash, reboot and post diagnostics.

 

I just went down, here is the log. I also saw a ton of "nvidia" on the screen before I rebooted the system. 

log_20250129.log

tower-syslog-20250205-1152.zip tower-syslog-20250205-1952.zip

Edited February 5, 20251 yr by tgotr

I said "diagnostics" not only syslog...

 

4 minutes ago, MAM59 said:

I said "diagnostics" not only syslog...

 

Ah, it crashed so hard, I couldn't SSH in.

ok, then there is little hope for you. my only guess would be that some pcie lanes are either shared or some (parts of) slots are deactivated.

The HBA becomes crazy if not all lanes are available (but they are only needed if some traffic is going on or certain addresses are used,  so it is not a problem all the time)

 

1 hour ago, MAM59 said:

ok, then there is little hope for you. my only guess would be that some pcie lanes are either shared or some (parts of) slots are deactivated.

The HBA becomes crazy if not all lanes are available (but they are only needed if some traffic is going on or certain addresses are used,  so it is not a problem all the time)

 

Yep, I think you're right. The system is running it in X4, and it's an X8 card. I tried to force the board to do x8-x8, and it refused to post. it posts in Crypto Mining mode, which apparently allows PCIe cards to address more memory above 4G

The way it sounds, I need to either drop in an APU, and ditch the 1650, or go back to 1x Sata controllers. 

EDIT
The cache disk where the system dumps itself on boot is on it's way out. I am in the process of moving everything off that cache pool and onto another. 

Edited February 5, 20251 yr by tgotr

Since you have a bunch of plugins or dockers installed, try running without them and see if it still crashes, the cause doesn't have to be the card. Also verify you have the latest firmware for the card (Build 32118).

On 2/6/2025 at 2:37 AM, Wody said:

Since you have a bunch of plugins or dockers installed, try running without them and see if it still crashes, the cause doesn't have to be the card. Also verify you have the latest firmware for the card (Build 32118).

I am running with Docker turned all the way off at the moment. I also took out the NVMe Drive that was reporting a ton of bad sectors. If anything, taking that off the PCIe bus should free up some lanes for both the GPU and HBA. 

I also have been trying to boot freeDOS to check the firmware on the card, and the board won't boot legacy boot items. I may have to pull the card and put it in another machine that can boot legacy and has PCIe unless I can find a way to do the upgrade with a linux live environment. 

13 hours ago, tgotr said:

I also have been trying to boot freeDOS to check the firmware on the card

Freedos works, linux won't, because the software is DOS and needs hardware-access. Apparently you can also use maxView storage manager and arcconf which may have Linux versions, but the software is from 2013, so I doubt it works any better. I used rufus with the included freedos.

58 minutes ago, Wody said:

Freedos works, linux won't, because the software is DOS and needs hardware-access. Apparently you can also use maxView storage manager and arcconf which may have Linux versions, but the software is from 2013, so I doubt it works any better. I used rufus with the included freedos.

Like I said, the board is unable to boot legacy BIOS systems. There is no option in the UEFI settings to allow it. It sucks. I do have a few machines that I can throw the HBA into, just to check the bios, so it's not a huge deal. 

It crashed again while writing up this reply, Docker not running at all. About 36 hours of uptime and out of nowhere, crash. 

There was some odd behavior behavior before this crash, it was slow to load directories on the server over the network.

I just updated the firmware, it was a few versions behind. 

I just got it to crash again, trying to move around network directories. It's likely what another user said, not enough lanes and when it tries to access the lanes it needs, it crashes the system.
 

Edited February 9, 20251 yr by tgotr

Well, that's not good.  3 out of 4 sticks are bad. 

Edited February 13, 20251 yr by tgotr

dont give up so early.

It is well known that many Boards cannot run the RAM with the desired speed.

Try to lower the frequency to 2400 and try again. DO NOT USE THE XMP SETTINGS of the RAMs, do not overclock!

 

Quote

 

Big news
 

I finally got the system to run stabally for more than a few hours.

I had bad ram sticks, failed testing at slower speeds. 3 our of 4 sticks went bad, but not in a way that kept the system from POSTing. With the RAM being bad, I moved the docker.img to one of my cashe pools and docker is stable. I pulled DDR4 sticks out of my main rig, and the ram sockets tested fine. 

I also identified a common cause of it crashing. When it accessed the drives on the 3rd and 4th connector on the HBA. It's less crashy on the 3rd connector, leading credence to the theory that running the card at 4X when it's an 8X card made it sad. I swapped some cables around and the crashing behavior occurred when accessing drives on the 3rd and 4th connector. It would work for a bit, but if I tried copying files from those drives, it would start at 200mb/s, then it would slow down and then crash the system. 

I am putting in new ram sticks and a SATA card that works on 4x.