1) Bring down applications.
2) Bring down databases.
3) Unmount all filesystems.
4) Convert all vpath VGs to hdisk VGs
a) lsvg -o | grep -v rootvg | xargs -n1 varyoffvg
b) rm /tmp/vp2hd.out
c) lsvg | grep -v rootvg | while read vg
do
vp2hd $vg >> /tmp/vp2hd.out
echo $vg
done
d) Check contents of /tmp/vp2hd.out. Ensure all VGs converted successfully.
e) lsvg -o | grep -v rootvg | xargs -n1 varyoffvg #
5) Remove all vpath devices and take dpo down to "Defined" state
a) lspv | grep vpath | awk '{print $1}' | xargs -n1 rmdev -dl
b) rmdev -l dpo
6) Remove all FC hdisk devices and fscsi drivers; take FC adapters to "Defined" state
a) lsdev -Cc driver | grep ^fscsi | awk '{print $1}' | xargs -n1 rmdev -dRl
b) lsdev -Cc adapter | grep ^fcs | awk '{print $1}' | xargs -n1 rmdev -l
7) Make your SAN change
8) Rediscover all hdisk devices
a) lsdev -Cc adapter | grep ^fcs | awk '{print $1}' | xargs -n1 cfgmgr -vl
# Or run this by hand one adapter at a time if you prefer
9) Verify all hdisk devices are present
a) lspv
b) lsdev -Cc disk
10) Fix queue_depth for all FC hdisks
a) lsdev -Cc disk | grep 2105 | awk '{print $1}' | \
xargs -n1 chdev -a queue_depth=20 -l
# NOTE: This will take a long time. We've seen ~45 minutes for about 600 disks
11) Recreate vpath devices
a) cfgmgr -vl dpo
12) Verify all vpath devices are correct
a) lsvpcfg
b) datapath query device
13) Convert all hdisk VGs to vpath VGs
a) rm /tmp/hd2vp.out
b) lsvg | grep -v rootvg | while read vg
do
hd2vp $vg >> /tmp/hd2vp.out
echo $vg
done
c) Check /tmp/hd2vp.out to ensure all VGs were converted successfully
# Volume groups have been left online by hd2vp
14) Mount all filesystems
15) Bring up databases.
16) Bring up applications
No comments:
Post a Comment