vgextend in a cluster

[derek b smith_1]

Hi!

I have version 11.14 of mcsg. I need to add a new LUN to an existing vg in an existing active cluster/package. Does the package need to come down?
I looked online and found this. Is this accurate for 11.14 hpux 11.11 online jfs?
thanks!


On Command View EVA
a)--Create Vdisk at required size
b)- Present Vdisk to the host
c)- do a ioscan, insf to make LUN's visible on both nodes;
d)--If you have autopath then do otherwise skip this step
#autopath discover
#autopath display
e)-- cmhaltpkg pkg (halt package, holding this group)
f) ---vgchange -c n vg01
g)-- vgchange -a y vg01
h)-- pvcreate /dev/rdsk/c10t3d8 ---------substitute the correct value
i) --vgextend vg01 /dev/dsk/c10t3d8 -----take care to add all the alternative links if you are using pvlinks but in case of autopath a single device file should take care this
j)-- lvextend -l 200 /dev/vg01/lvol1
k)-- extendfs /dev/vg01/lvol1
l)- export in PREVIEW mode volume group and create map file of the group;
- on the other node(s) export same group, recreate /dev/vg01/group node and import group from map file.
m)-- vgchange -a n vg01
n)---vgchange -c y vg01
0)-- cmmodpkg -e -v pkgname


http://docs.hp.com/en/ha.html#Serviceguard

[Johnson Punniyalingam · Accepted solution]

>>Does the package need to come down?

It depends (if you want "cluster aware logical volume used immediate)you need to "cmhaltpkg" & "cmrunpkg" or if you have any planned down time you can do also on that time.

performing "LVM" activities, doesn't really require pkg downtime. Only thing is newly created cluster aware logical volume will be can be manually mounted.

Once disk presented to host,

# pvcreate /dev/rdsk/cxtxdx

# vgextend /dev/vgname /dev/dsk/cxtxdx

# lvcreate -L /dev/vgname/lvname

# newfs -F vxfs -o largefiles /dev/vgname/rlvname

# cp -p /etc/cmcluster/pkgname/pkg.cntl /etc/cmcluster/pkgname/pkg.cntl.org

# vi /etc/cmcluster/pkgname/pkg.cntl

LV[33]=/dev/vgname/lvname; FS[33]=/oracle/ABN/sapdata20; FSMOUNT_OPT[33]="-o convosync=direct,mincache=direct,delaylog,nodatainlog"

(Add newly create Logical volume to pkg.cnt.file)

# vgexport -p -s -v -m /tmp/vgname.map /dev/vgname

# rcp /tmp/vgname.map nodee2:/tmp/vgname.map

# cp -p /etc/cmcluster/pkgname/pkg.cntl /etc/cmcluster/pkgname/pkg.cntl.node2

# rcp /etc/cmcluster/pkgname/pkg.cntl.node2 nodee2:/tmp/


(mount the mount mpoint manually on the node where the vg is active now

# mount -o convosync=direct,mincache=direct,delaylog,nodatainlog /dev/vgname/lvolname /oracle/ABN/sapdata20


On Adoptive Node :-

# vgexport /dev/vgname

# mkdir /dev/vgname

# ll /dev/vg*/group

# mknode /dev/vgname/group c 64 0x0X0000

# vgimport -p -s -v -m /tmp/vgname.map /dev/vgname -> (preview) if no error proceed below by removing -p

# vgimport -s -v -m /tmp/vgname.map /dev/vgname

# cp /tmp/pkg.cntl.node2 /etc/cmcluster/pkgname

# cp -p /etc/cmcluster/pkgname/pkg.cntl /etc/cmcluster/pkgname/pkg.cntl.org

# mv /etc/cmcluster/pkgname/pkg.cntl.node2 /etc/cmcluster/pkgname/pkg.cntl

HTH,
Rdgs,
Johnson

[Kapil Jha]

Yups no need to halt the package you can manually mount it and next time as you have edited the config files it would be cluster aware and would be mounted when pkg would start.

Just do LVM things normally, and do not forget the map file.

BR,
Kapil+

[Chandrahasa s]

Hi,
Do you have online JFS ,if you have you can extend vg without downing the cluster or package.

[derek b smith_1]

Its a pre-existing active NFS package, local as vxfs.
I want to vgextend vg01, then lvextend then fsadm to 120Gb.

vg01 is now in available,exclusive mode.
can i vgextend this without taking the package offline, and the vg taken away from cluster?

thanks for the clarification!

[Nath_3]

Hi Derek,

Since this a vgextend you need to bring down the package . I think this is the safe.

Johnson, has given the correct steps.

Thanks
Ravi