[6.9.2] HBA card backplane startup reliability

[Sim1]

So I recently upgraded my server to use a backplane (specifically a supermicro BPN-SAS2-846EL1). I am running an IBM 1015 card flashed to IT mode. I previously had been using a SAS to sata cable for my drives, but now have the backplane connected with two of the SAS-SAS cables. 

 

The issue I am having is that since switching to use the backplane I am having some start up issues. Essentially, it seems to be a dice role as to whether the HBA card/backplane get recognized by unraid on boot up. If it is not recognized, then essentially unraid will show all drives as missing, however, if it is recognized, then it works perfectly, including hotswap features, and read/write speeds are excellent. 

 

I'm not really sure exactly what to make of it. Since the backplane is working well when recognized, I would assume that the issue is more likely to be related to the HBA, and I'm not sure if I need to get a new HBA, or whether it could be something simpler, like the cable configuration is wrong (I have both cable slots for the card connected to J0 and J1 on the backplane). 

 

Either way, it's kind of frustrating because it means anytime I do a reboot right now, I have to deal with the possibility of troubleshooting the server for a while, just rebooting it until the backplane is properly recognized, after which it will be extremely stable. 

 

Thanks for any insight on this in advance. 

[trurl]

attach diagnostics to your NEXT post in this thread

[Sim1]

Here are the diagnostics

mediaserver-diagnostics-20220501-2056.zip

[JorgeB]

Unfortunately not easy to diag without using a different HBA or a different backplane, cable config is OK and doubt it's cable related, but you can try with just one cable connected at a time to rule that out, bandwidth will be reduce but it's fine for testing.

[JorgeB]

You might still post diags when it doesn't find any disks just in case there's some issue initializing the HBA.

 

P.S. disk10 is failing in case you didn't notice yet.

[Sim1]

OK I'll post logs when the card fails to be recognized on startup. 

 

I know about disk 10. Getting a replacement drive has been a bit of a saga including one drive from Amazon with a bogus warranty that I had to return and another drive that failed preclear...

[Sim1]

Ok so a little update here - no more logs, but I think I found the culprit, though I am not entirely sure why it is occurring. 

 

So I also use my unRAID server to host a zwave docker, and the server is not in an optimal position for the z wave stick, so I have had it connected to the server via usb over ethernet. The USB over eth female end has a power supply, which appears to supply 5 V power to allow for USB to be transmitted over what would normally be too long of a run for USB (or so I understand). I believe that this power supply was the culprit for a couple of reasons. 

 

1. Whenever the server would shutdown, it would cause an odd continuous alarm sound, that would not resolve with unplugging both of the power supplies (the new server chassis came with a redundant PSU) and would only resolve when removing the implicated USB. 

 

2. [Most Importantly] When the system would start up with the USB connected, the system would not recognize the drives. When the usb was disconnected, then it would seem to recognize all of the drives without issue. 

 

Also interestingly, when I disconnected the power supply part of the USB over ethernet, the USB device still worked normally. 

 

My theory is that there was something from the USB power supply that was essentially backfeeding into the motherboard in some fashion that affected the HBA card enough to prevent it from recognizing the drives. 

 

I don't have some massive insight into the issue, but only wanted to post this as a potential solution to people who might be having the same issue down the road. 

[JonathanM]

54 minutes ago, Sim1 said:

and the server is not in an optimal position for the z wave stick, so I have had it connected to the server via usb over ethernet.

There is a way to do USB over ethernet, such that each end has an IP address and can pass through routers, but I forget where I saw it. What you have, and what I run as well, is USB over CAT5 point to point. It's really just a pair of dumb transceivers, that attempts to clean up and impedance match the signal enough to use over a long run of network cable. The adapters I have don't have a provision for power, so what I did was make sure to have a powered USB hub at one or both ends of the run, depending on what I am connecting. I have runs to 3 different spots in the house with monitors, keyboards, and mice. The HDMI is split with a monoprice blackbird amplifier, and all the keyboards and mice feed back to a single VM. That way I can play media or browse simultaneously all over the house.

 

If your 5v adapter was at the zstick end, it's possible that the ground leg on that circuit wasn't bonded properly to the circuit at the server end, causing a constant trickle of power through the whole USB circuit. Or the 5v adapter was poorly designed or faulty. In any case, if it runs fine without the power supply, great, but if you start getting glitches, try using a powered USB hub at the server end or the device end instead of the adapters native supply and see if it helps.