Raid 5 on Fedora 4 working with SATA ?

[Terry Barnaby]

Gilboa Davara wrote:
> On Wed, 2006-02-01 at 12:01 +0000, Terry Barnaby wrote:
> 
>>Gilboa Davara wrote:
>>
>>>By default software RAID1/5/6 support on-line drive
>>>kill/remove/rebuild/etc.
>>>However, seems that the MD driver is unaware of the dead drive.
>>>
>>>What does /proc/mdstat say?
>>>
>>>Gilboa
>>>
>>>
>>
>>After removing the SATA cable on /dev/sdd, if I access a file there is a long delay
>>and then the program returns with no error but no data. For example:
>>"cat /data/test-file" will delay and then exit with status of "0" but no file
>>contents are displayed.
>>
>>The kernel is: 2.6.14-1.1656_FC4smp: I get the following kernel messages:
>>
>>Feb  1 11:51:37 library kernel: ata2: command 0x35 timeout, stat 0x0 host_stat 0x61
>>Feb  1 11:51:38 library sshd(pam_unix)[13027]: session opened for user root by root(uid=0)
>>Feb  1 11:52:07 library kernel: ata2: command 0x25 timeout, stat 0x0 host_stat 0x61
>>Feb  1 11:53:07 library last message repeated 2 times
>>Feb  1 11:54:37 library last message repeated 3 times
>>Feb  1 11:55:01 library crond(pam_unix)[13091]: session opened for user root by (uid=0)
>>Feb  1 11:55:01 library crond(pam_unix)[13091]: session closed for user root
>>Feb  1 11:55:07 library kernel: ata2: command 0x25 timeout, stat 0x0 host_stat 0x61
>>
>>/proc/mdstat has:
>>Personalities : [raid1] [raid5]
>>md1 : active raid1 sdc1[0]
>>       20482752 blocks [2/1] [U_]
>>
>>md2 : active raid5 sdd3[3] sdc3[2] sdb3[1] sda3[0]
>>       873196800 blocks level 5, 64k chunk, algorithm 2 [4/4] [UUUU]
>>
>>md0 : active raid1 sdb1[1] sda1[0]
>>       20482752 blocks [2/2] [UU]
>>
>>unused devices: <none>
>>
>>The output of "mdadm -Q --detail /dev/md2" is:
>>/dev/md2:
>>         Version : 00.90.02
>>   Creation Time : Tue Jan 31 14:14:07 2006
>>      Raid Level : raid5
>>      Array Size : 873196800 (832.75 GiB 894.15 GB)
>>     Device Size : 291065600 (277.58 GiB 298.05 GB)
>>    Raid Devices : 4
>>   Total Devices : 4
>>Preferred Minor : 2
>>     Persistence : Superblock is persistent
>>
>>     Update Time : Wed Feb  1 11:51:07 2006
>>           State : active
>>  Active Devices : 4
>>Working Devices : 4
>>  Failed Devices : 0
>>   Spare Devices : 0
>>
>>          Layout : left-symmetric
>>      Chunk Size : 64K
>>
>>            UUID : 56bd5037:9d9b9018:eb8f01d6:94155776
>>          Events : 0.230
>>
>>     Number   Major   Minor   RaidDevice State
>>        0       8        3        0      active sync   /dev/sda3
>>        1       8       19        1      active sync   /dev/sdb3
>>        2       8       35        2      active sync   /dev/sdc3
>>        3       8       51        3      active sync   /dev/sdd3
>>
>>Terry
> 
> 
> Very weird.
> I've got a number of both IDE, SATA and SCSI RAID5 setups and I never
> seen such a problem.
> What happens if you try to access the RAID5 array?
> (hdparm -tT /dev/md2)
> 
> Gilboa
> 

I hav'nt tried "hdparm -tT /dev/md2", but if I access a file there is a long delay
and then the program returns with no error but no data. For example:
"cat /data/test-file" will delay and then exit with status of "0" but no file
contents are displayed.

This is VERY VERY BAD !

I really think this must be a bug, possibly in the SATA driver in Fedora Core 4's
2.6.14-1.1656_FC4smp kernel. I have a spare SCSI system with 3 SCSI disks,
I will set that up and see how this handles the situation ...

Terry