Kanoa Withington wrote:
Ideally a different HBA altogether, but a different channel on a multichannel HBA at a minimum. If your SCSI card is not a multichannel card, think about getting one or think about a completely different arrangement.
It may be possible to tune the HBA reset behavior or the XFS timeout
threshold but as a matter of principle when constructing disk mirrors
you should try to keep the disks as separate as possible. You should
only need to tune, tweak or patch if you are trying to do something
unusual - which you are not.
Very true.
The default parameters for SCSI (5 retries as I recall) can take a very long time when a SCSI bus reset is called for (settle times and such) - I've seen 2+ minutes. Even with totally redundent controllers a logical I/O (to the RAID) could be held up waiting for a physical I/O by this long. The XFS parameter would need to be raised above the threadhold.
mark
In the short term, unplug the failing disk:
Jan 10 11:56:06 linux-sg2 kernel: SCSI disk error : host 0 channel 0 id 0 lun 47
You are better off without it if your system is crashing.
-Kanoa
On Thu, 20 Jan 2005, David Dougall wrote:
-By "different controller" do you mean HBA controller or disk controller? The disk devices are on completely different jbods. They are both through the same HBA(the server only has 1 PCI slot) --David Dougall
On Thu, 20 Jan 2005, Kanoa Withington wrote:
Yes, that's a standard XFS timeout and shutdown. If your second disk is on the sme SCSI channel try moving it to a different one, preferably a different controller alotgether.
Your disk 08:10 does have real problems, but they are separate from the XFS shutdown which should be prevented by the MD layer.
-Kanoa
On Thu, 20 Jan 2005, David Dougall wrote:
return code = 8000002 Jan 10 11:56:08 linux-sg2 kernel: Info fld=0xc7c0181, Current sd08:10: sense key Hardware Error Jan 10 11:56:08 linux-sg2 kernel: I/O error: dev 08:10, sector 209453441 Jan 10 11:56:08 linux-sg2 kernel: I/O error in filesystem ("device-mapper(254,1) ") meta-data dev device-mapper(254,1) block 0x18fa318f ("xlog_iodone") err or 5 buf count 2048 Jan 10 11:56:08 linux-sg2 kernel: xfs_force_shutdown(device-mapper(254,1),0x2) c alled from line 966 of file xfs_log.c. Return address = 0xc0246d9b Jan 10 11:56:08 linux-sg2 kernel: Filesystem "device-mapper(254,1)": Log I/O Err or Detected. Shutting down filesystem: device-mapper(254,1) Jan 10 11:56:08 linux-sg2 kernel: Please umount the filesystem, and rectify the problem(s)
I don't see any error messages from md in any of these logs. --David Dougall
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
