I am sending this to the afshelp desk as well, but its the same message to all.
Hi all, does anybody out here have experiences with h/w raid disk and afs vice
partitions. Here is my config. I have a dual controller nstor(andataco) raid
with a failover driver in software running on a Sun ultra 60 with 2.6 os and
afs 3.5 . I have read/write io in progress. When we have or cause a fail of
the controller path on a channel say, channel 2 the drives move logically to
channel 1. Things move nicely from a unix perspective. I can see the vice
partitions and there contents that where on channel 2. The fileserver process
cored dumped and restarted and would not attach the volumes that would have
been on channel 2. The afs volumes on channel 1 were ok. It reported in
filelog that the volumes would not attach and a vfsck was needed. A salvage of
the volume or partition had the same result. The problem was the the salvage
or vfsck could not access the disk. I tried a simple dd if=/dev/dev/c2t5d0s6
of=/dev/null and was able to read it fine. So I could see them from unix. So I
did the command to move the disks back to the correct controller and the
drives moved fine as far as unix was concerned. Afs was still not happy. I
salvaged the volumes and now things where back to normal. So this time afs
could see the drives. I did not have to do a vfsck nor did I try.
So while the drives where on the alternate path, the afs utilites could not
see the disks, in particular the hidden parts (inodes and real data) on the
/vicepn. When on the primary path afs could see the /vicepn's.
Thanks for any input. Looking to here from transarc.
Stan fermi-lab.
