April 11, 201511 yr Got an issue with a new-build server into which I've moved an existing 22-disk array plus cache. On start up and intermittently thereafter I'm getting the following posted in the syslog. Tower2 kernel: ata1.00: exception Emask 0x10 SAct 0x0 SErr 0x280100 action 0x6 frozen (Errors) Tower2 kernel: ata1.00: irq_stat 0x09000000, interface fatal error (Errors) Tower2 kernel: ata1: SError: { UnrecovData 10B8B BadCRC } (Errors) Ata1 (and it's always ata1 that has the error) is associated with the disk in slot 3 of the chassis - that is, the rightmost slot on the top row. All the forum info I have seen so far indicates that Bad CRC errors are caused by hardware issues, power, cables etc. So this is what I've done so far: 1. Put half the backplanes (3) on a second 12V power rail 2. Replaced the SATA cable with a known good one from another chassis (twice) 3. Moved the individual cable for slot 3 to another SATA socket on the motherboard 4. Swapped over the top two backplanes in the chassis No matter what, the errors return and are always pointing to slot 3. My next thought was the disk itself so I ran a short SMART test which it passed and have just started a long one which I'll post later. There are no problems reported on the MyMain Smart page. In the meantime does anyone have an insight into what might be causing this? Or even whether this is anything to be concerned about? The array itself seems to working fine and IPMI confirms that the CPU temperature is remaining below threshold. Configuration: Norco 4224, Supermicro X10SL7-F, 2x AOC-SASLP-MV8, Core i7, 16GB and Corsair 850W psu. Running Unraid 5.0.5. syslog4_-_changed_backplanes.zip smart-short2.txt
April 12, 201511 yr Could be a bad SATA port. Switch parts and test until the bad component is identified.
April 12, 201511 yr Author Yes, I tried that in #3 above. The motherboard has 14 SATA ports so that's an easy one. Also put the backplane (which supports 4 disks) on a separate power rail, but no change. Presently running another long smart test since the only constant in all this appears to be the drive in slot 3. Strangely though the system is telling me it will take 520 minutes, the previous one took 255.
April 12, 201511 yr Yes, CRC errors are almost always related to the cable or motherboard SATA port. The other possibility is electronic noise. Are your SATA cables bundled? Is that particular cable running near a possible source of noise that no other SATA cables are near? That said, based on the troubleshooting you've done above, the only common factor is indeed the disk. Possibly the SATA connector on the drive itself has issues. As you've done, I'd do a long SMART test and see what the results are. The only other thing I could think of to do (that you haven't already done) would be to rebuild the affected drive with a replacement drive and see if the issue continues. Run a couple pre-clear passes on the pulled drive and see what the results are.
April 13, 201511 yr Author Well the mystery deepens. The disk passed the long smart so I replaced it with a newly pre-cleared one and let the system start the rebuild. Everything seemed to be OK at first but then several hours into the process there was a flurry of errors including some '10B8B LinkSeq' that I hadn't seen before. All again pointing to ata1 and slot 3. Right now I'm planning to let the rebuild finish and then move the disk to the slot currently occupied by the cache leaving slot 3 empty.
April 14, 201511 yr 10B8B errors are very specific to the SATA hardware path. Most likely a bad SATA cable or port.
April 14, 201511 yr Author Ok thanks. Once the parity check completes I'll go through the cable/port routine again to make sure I didn't miss anything.
Archived
This topic is now archived and is closed to further replies.