Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Running Debian 7, mdadm 3.2.5. I issued a command to grow a 5 disk, 3 spare raid 5 array to an 8 disk, 0 spare raid 6 array:

mdadm --grow /dev/md0 --level=6 --raid-devices=8 --backup-file=/root/raid5backup

Between one and five seconds after the drives spin up, there is a little clikc sound and they stop, and a ton of error messages spit out to the console. Googling chunks from these messages (there's more every 1-5 seconds so I can't get it exactly) finally landed me with this libata error message breakdown page.

According to that linked page, my issue is "communications between device and host temporarily lost, but regained". Looks like a bus error - could it be a shoddy cable? This machine has 11 SATA devices, 6 through onboard connectors and 5 more split over two expansion cards. The motherboard is brand new.

I'm not sure what to do with the messages though. Should I shutdown and try replacing cables? Any ideas?

share|improve this question

1 Answer 1

up vote 3 down vote accepted

My theory is that the power supply to the drives can't handle the load. When you issue the grow order, all the drives in the array have to start doing real work. This consumes more power. The power supply can't deliver and the voltage to the drives dips. Once the voltage gets too low, their motors shut off briefly - causing the clicking sound as their heads get moved to safety position, and their logic controllers lose communication with your SATA controller - causing the errors. When the motors shut off, the load on the power supply is reduced, and the voltage it puts out recovers to normal levels. So the drives start back up and the cycle repeat.

You could try another power supply, removing a couple drives from the system (not easy I know) or removing other sources or moving them externally (for example, maybe attach fans to another power supply.) Or, you could try rebalancing the drives to different cables from the power supply. (You could also check the power supply's ratings and the ratings of the attached devices and see if you are in safe range.)

share|improve this answer
    
It seems like 530 watts should be enough for 8 HDDs, 2 SSDs, 7 case fans, a new Xeon and an ancient PCI video card, but you're right. The case was open, so I (very carefully!) unplugged the spare SSD and 4 of the fans. After the last fan was out, it the noises and libata errors stopped and the rebuild time went from 2000min to 750! All the ata errors were listed as coming from the "rebuilding" drive in mdadm --detail, so I guess that's where all the writes are happening and therefore the largest power drain. Thanks buddy, you're a life saver! –  ACK_stoverflow Nov 14 '13 at 6:15
    
Glad it worked out!! –  whitepaws Nov 14 '13 at 6:46

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.