topic Re: Migration Question for the VxVM hotshots in Operating System - HP-UX

Migration Question for the VxVM hotshots

S.Rider — Tue, 11 Apr 2006 11:25:38 GMT

We have a pair of 11.11 servers running HP's VxVM 3.5.
Server-1 has 150+ veritas volume groups containing oracle data. It also has a rootdg which also contains 100gb of oracle data.
Server-2 has a small empty rootdg and that's it.
The plan is to shutdown both servers, re-assign the XP12000 LUN's from server-1 to server-2, and the 1 rotdg lun from server-2 to server-1, reboot the 2 servers, and Veritas will "magically" find the moved volume-groups (including rootdg) and come up with everything. Or at least, that's what one group is telling us. There's a dissenting voice (that I tend to believe) who is saying this will work fine for all but rootdg.
Has anyone actually done something like this ?
Anyone think it will work for rootdg ?

Re: Migration Question for the VxVM hotshots

sathish kannan — Wed, 12 Apr 2006 06:47:46 GMT

Hello Rider,

What about the host name? Are you going to transfer them? because while importing dg to other nodes veritas looks for the hostname.

Regards
Sathish

Re: Migration Question for the VxVM hotshots

Robert-Jan Goossens — Wed, 12 Apr 2006 07:16:31 GMT

Yes it will if your hostname and ip-address are the same. In fact the methode you are using is my DR concept.

with version 3.5, you don't need to export your current diskgourp, as soon as the server boots it will detect/import the diskgroup automaticly.

reassign your LUN --> can you add the wwn's of the second host HBA, if so you don't need to reassign your lun.

As for your question about your rootdg, if you have some space left in one of your diskgroups I would copy it, otherwise I would create a backup and restore the data to a new filesystem in the correct diskgroup. You dont want to merge and importing data at the same time.

Hope this helps,
Robert-Jan

Re: Migration Question for the VxVM hotshots

S.Rider — Wed, 12 Apr 2006 19:23:32 GMT

Hostname and IP-Address will not be changed on either server.They have to remain as they were ven after the other servers data-groups are restored to it.

Re: Migration Question for the VxVM hotshots

chris huys_4 — Thu, 13 Apr 2006 04:29:42 GMT

Hello Rider,

>We have a pair of 11.11 servers running HP's
>VxVM 3.5.
>Server-1 has 150+ veritas volume groups
> containing oracle data. It also has a
> rootdg which also contains 100gb of oracle > data.
> Server-2 has a small empty rootdg and
> that's it.

Has anyone actually done something like this ?
yes

Anyone think it will work for rootdg ?
yes.

NOTE : You will need with this action plan to reinstall the OS on youre server2 on the same lun that you are using now to boot server2 for or a seperate lun.

actionplan :
0. on server1

0.a
collect output of
#vxdisk -g rootdg list
#vxdisk list

0.b. shutdown server1
#shutdown -h 0

1. on server2

1.a Install core-os using LVM as volumemanager.
[pretty straightforward]

1.b start vxconfigd in disabled mode.
Booted from youre just installed disk :
# vxconfigd -km disable
vxvm:vxconfigd: NOTICE: Setting max thread limit to 60

1.c. create the /etc/vx/volboot file
# vxdctl init

1.d. create a "empty" rootdg
# vxdg init rootdg

1.e. check if vxdisk list works and check if the rootdisk devices of server1 are "listed"

# vxdisk list
DEVICE TYPE DISK GROUP STATUS
c0t0d0 simple - - online
c0t1d0 simple - - online
c0t2d0 simple - - LVM
c0t3d0 simple - - LVM
c2t0d0 simple - - online
c2t2d0 simple - - online
c4t2d0 simple - - LVM
c4t3d0 simple - - LVM
c4t4d0 simple - - online
c4t5d0 simple - - online
c4t6d0 simple - - LVM
c4t7d0 simple - - online

1.g. check 1 of the rootdisks
We know from before that rootdg consisted of the disk c2t0d0/c0t1d0 and spare disk c2t2d0

dm rootdisk01 c2t0d0 simple 1024 35556864 -
dm rootdisk02 c0t1d0 simple 1024 35556864 -
dm rootdisk03 c2t2d0 simple 1024 8889105 SPARE

# vxdisk list c2t0d0
Device: c2t0d0
devicetag: c2t0d0
type: simple
hostid: loopback
disk: name= id=1091461034.1039.loopback
group: name=rootdg
id=1091461034.1025.loopback
flags: online ready private autoconfig
autoimport
pubpaths: block=/dev/vx/dmp/c2t0d0
char=/dev/vx/rdmp/c2t0d0
version: 2.2
iosize: min=1024 (bytes) max=256 (blocks)
public: slice=0 offset=3168 len=35556864
private: slice=0 offset=2144 len=1024
update: time=1091654242 seqno=0.40
headers: 0 248
configs: count=1 len=727
logs: count=1 len=110
Defined regions:
config priv 000017-000247[000231]: copy=01 offset=000000 enabled config priv 000249-000744[000496]:
copy=01
offset=000231 enabled
log priv 000745-000854[000110]: copy=01 offset=000000 enabled lockrgn priv 000855-000919[000065]:
part=00 offset=000000
Multipathing information:
numpaths: 1
c2t0d0 state=enabled

>> c2t0d0 shows that the hostid is loopback

1.h. change hostid to loopback
# vxdctl init loopback

# vxdctl list
Volboot file
version: 3/1
seqno: 0.7
cluster protocol version: 40
hostid: loopback
entries:

1.i. add the disks of rootdg to the volboot file

# vxdctl add disk c2t0d0
# vxdctl add disk c0t1d0
# vxdctl add disk c2t2d0

# vxdctl list
Volboot file
version: 3/1
seqno: 0.10
cluster protocol version: 40
hostid: loopback
entries:
disk c2t0d0 type=simple
disk c0t1d0 type=simple
disk c2t2d0 type=simple

1.j. "reset" and "enable" vxconfigd daemon

# vxconfigd -kr reset -m boot
vxvm:vxconfigd: NOTICE: Setting max thread limit to 60
vxvm:vxconfigd: WARNING: Disk c4t7d0 names group rootdg, but group ID differs

# vxconfigd -km enable
vxvm:vxconfigd: NOTICE: Setting max thread limit to 60

10. check if rootdg is imported
# vxdisk list
DEVICE TYPE DISK GROUP STATUS
c0t0d0 simple - - online
c0t1d0 simple rootdisk02 rootdg online
c0t2d0 simple - - LVM
c0t3d0 simple - - LVM
c2t0d0 simple rootdisk01 rootdg online
c2t2d0 simple rootdisk03 rootdg online spare
c4t2d0 simple - - LVM
c4t3d0 simple - - LVM
c4t4d0 simple - - online
c4t5d0 simple - - online
c4t6d0 simple - - LVM
c4t7d0 simple - - online

>> CONCLUSION -- rootdg is imported, which contains the disks we previously booted from, on a newly installed system booted from lvm.

Good Luck,
Chris