Bug 344717

Summary:	sys-kernel/gentoo-sources-2.6.34-r12: software raid fails/drops every sata drive in array, but drives okay after reboot (sas)
Product:	Gentoo Linux	Reporter:	dak <admin>
Component:	[OLD] Core system	Assignee:	Gentoo Kernel Bug Wranglers and Kernel Maintainers <kernel>
Status:	RESOLVED NEEDINFO
Severity:	critical	CC:	alexanderyt, pchrist
Priority:	High
Version:	unspecified
Hardware:	All
OS:	Linux
Whiteboard:
Package list:		Runtime testing required:	---
Attachments:	dmesg output from first resync failure lspci -v, affected array is attached to marvell sas controller cat /proc/mdstat kernel .config dmesg from 2.6.36 kernel resync successful dmesg from second machine fails all drives in the array

Description dak 2010-11-08 18:39:31 UTC

This happens with gentoo-sources-2.6.34-r12. Software raid5 array fails to resync because all drives are dropped from the array. 

Reproducible: Always

Steps to Reproduce:
1. echo check > /sys/block/md2/md/sync_action
2.
3.

Actual Results:  
All drives are dropped from the affected array md2. Will attach dmesg output.

Expected Results:  
Drives are only marked faulty if they are, in fact, faulty.

Comment 1 dak 2010-11-08 18:54:53 UTC

Created attachment 253673 [details]
dmesg output from first resync failure

Comment 2 dak 2010-11-08 19:01:18 UTC

Created attachment 253675 [details]
lspci -v, affected array is attached to marvell sas controller

Comment 3 dak 2010-11-08 19:01:52 UTC

Created attachment 253677 [details]
cat /proc/mdstat

Comment 4 dak 2010-11-08 19:06:13 UTC

Created attachment 253681 [details]
kernel .config

Comment 5 dak 2010-11-08 19:35:39 UTC

Created attachment 253691 [details]
dmesg from 2.6.36 kernel resync successful

Have changed kernels to gentoo-sources 2.6.36 and a resync was successful, but dmesg is still showing some similar errors.

Comment 6 dak 2010-11-08 20:13:58 UTC

Created attachment 253697 [details]
dmesg from second machine fails all drives in the array

Different machine, same result of dropped drives. This was not during a resync, just a sustained read from the array. I'm not sure about this one, it looks like the network died first? Anyway both the network and all drives in the raid array were gone, but were there after a reboot. The r8169 nic and the kernel were new, I had upgraded the kernel thinking the problems with the other machine were limited to mvsas, so either it's related or I'm just especially unlucky. I plan to change this one back to the very,very old kernel I was using previously.

Comment 7 dak 2010-11-09 15:00:15 UTC

(In reply to comment #6)
> Created an attachment (id=253697) [details]
> dmesg from second machine fails all drives in the array
> 
I found a flaky connection on this second machine.. still not sure how it affected all drives though.

Comment 8 Panagiotis Christopoulos (RETIRED) gentoo-dev

2010-11-20 20:27:48 UTC

Are you sure this isn't happening due to hardware problems? Are your drives OK (check them with smartctl (emerge smartmontools)? I assign this bug to the kernel team, as they probably will understand more from the dmesg logs etc.

Comment 9 Mike Pagano gentoo-dev

2010-11-20 22:56:29 UTC

Is this still an issue?