Corrupting SDS replicas on reboot
2007-12-25 10:35:00
My only reply was from shobhit@mindless.com. Shobhit mentioned that SSA
drivers need to load before SDS drivers, which makes sense, but doesn't
seem relevant since we have no problems with the SSAs.....
My mention of diskboards may have confused the issue. The diskboards are
essentialy just a means of mounting normal FW SCSI disks in a system that
has no other way of mounting normal SCSI disks.
The placement of additional replicas is the most important use of these disks.
I have rebuilt these slices, this time selecting 2MB (2 cylinders). It
didn't fix
the problem, and Sun now say to have at least 3 cylinders. This seems odd as
a replica only needs about 600KB (I'm told).
Since we can't reboot "willy-nilly" it would be nice to know the specific
requirement. I will try Sun's latest suggestion and let you know how it
turns out.
Meanwhile any futher info would be appreciated... ;-)
My post:
We have two E5000's configured almost identical:
Solaris 2.5.1
2 x SSA
1 x 4GB diskboard
Solstice DiskSuite 4.1
Patched to recomended set as of January this year.
The issue is with replica databases on the diskboard.
For those not familiar, the diskboard has 2 x 2.1GB disks. This slides into
the
chassis like a CPU board and then plugs into one of the FW SCSI ports. The
only reason we have this board is so that we have a third controller with
additional replicas to break the "stalemate" should a SSA be down and the
system is rebooted.
I partioned these drives with a 1mb on slice 6 and the rest on slice 0.
I then used metadb -a to add a replica to each disk on slice 6. That worked
fine until the system rebooted.
I found that the new replicas had gone into maintenace. So, I recreated them
this time I noticed from metadb -i that these replicas were not patched
into the
kernel.
With some repeated attempts I got the patch to occur. However on the next
reboot the replicas were in maintenance again, this time with a "M" signifying
start block problems! :-(
Sun have not given me a answer for this and have suggested allocating more
cylinders to slice 6.
Can anyone help?
--
David Robson
robbo@box.net.au
Davtin Systech Pty Ltd
Comments
Got something to say?
You must be logged in to post a comment.
