View previous topic :: View next topic |
Author |
Message |
sumrz n00b
Joined: 04 Feb 2004 Posts: 5 Location: Near Redmond, WA
|
Posted: Wed Mar 24, 2004 5:13 am Post subject: RAID5 corrupting partition table |
|
|
I am having trouble with my RAID5 on my SPARC machine corrupting the partition tables on the RAID drives. I also have a RAID5 on an Intel machine with IDE drives which has no problems. Both are set up similarly.
The machine is an Ultra2 Enterprise with 512MB RAM. The array is running off of a PTI sbus card and are the only set of drives on that SCSI chain. After I set up the partition tables, I can create the array, but it corrupts the partition table when I create a filesystem on it.
/etc/raidtab looks like this:
Code: | raiddev /dev/md1
raid-level 5
persistent-superblock 1
chunk-size 32
nr-raid-disks 6
nr-spare-disks 1
parity-algorithm left-symmetric
device /dev/scsi/host2/bus0/target0/lun0/part1
raid-disk 0
device /dev/scsi/host2/bus0/target1/lun0/part1
raid-disk 1
device /dev/scsi/host2/bus0/target2/lun0/part1
raid-disk 2
device /dev/scsi/host2/bus0/target3/lun0/part1
raid-disk 3
device /dev/scsi/host2/bus0/target8/lun0/part1
raid-disk 4
device /dev/scsi/host2/bus0/target9/lun0/part1
raid-disk 5
device /dev/scsi/host2/bus0/target10/lun0/part1
spare-disk 0 |
and the partitioning looks like this on all drives:
Code: | Disk /dev/scsi/host2/bus0/target10/lun0/disc (Sun disk label): 18 heads, 170 sectors, 5768 cylinders
Units = cylinders of 3060 * 512 bytes
Device Flag Start End Blocks Id System
/dev/scsi/host2/bus0/target10/lun0/part1 0 5735 8774550 fd Linux raid autodetect
/dev/scsi/host2/bus0/target10/lun0/part2 u 5735 5768 50490 82 Linux swap
/dev/scsi/host2/bus0/target10/lun0/part3 0 5768 8825040 5 Whole disk |
I'm hoping it is just a configuration error on my part. Any help is appreciated. |
|
Back to top |
|
|
sumrz n00b
Joined: 04 Feb 2004 Posts: 5 Location: Near Redmond, WA
|
Posted: Mon Mar 29, 2004 12:28 am Post subject: |
|
|
After no answer, apparantly nobody has any idea. I decided to rebuild the partition tables with the first partition starting at the 10th block. Not really a fix, but it seems to keep the tables intact now.
If anyone has any explaination, I'd love to hear it.
Thanks! |
|
Back to top |
|
|
thesimp n00b
Joined: 05 Jul 2004 Posts: 5
|
Posted: Sat Jul 10, 2004 9:20 pm Post subject: |
|
|
Hmm, I got the similar problem here on my Enterprise 250. Got a raid 5 setup with 3 18GB disks, no spare disk. All disks partitioned via fdisk with a slice 3 "Whole disk" and a slice 1 "Linux RAID auto". mkraid works ok, configuring a file system also works ok (mke2fs -j -b 4096 -R stride=8 /dev/md0), but after a reboot the partition table is destroyed on 1 or 2 disks. Error message "unknown partition table" appears in /var/log/messages.
Strange is that this only happens on /dev/sdc1 and/or /dev/sdd1. Never on the /dev/sdb1 disk.
I will try the workaround of starting the raid partition at blok 10. |
|
Back to top |
|
|
thesimp n00b
Joined: 05 Jul 2004 Posts: 5
|
Posted: Sat Jul 10, 2004 10:05 pm Post subject: |
|
|
As a followup to the previous post: it seems that starting the Linux RAID auto partition at block 10 instead of 0 is indeed the workaround. |
|
Back to top |
|
|
spam_ Tux's lil' helper
Joined: 14 Jan 2004 Posts: 105 Location: /dev/null
|
Posted: Thu Jul 15, 2004 6:37 pm Post subject: |
|
|
thesimp wrote: | As a followup to the previous post: it seems that starting the Linux RAID auto partition at block 10 instead of 0 is indeed the workaround. |
Or block 1. I was banging my head for a while with a 2-disk RAID-0 because of this. If you're using RAID (at least on sparc) you can't start at block 0 because the partition table and boot area get trashed. |
|
Back to top |
|
|
|