1834772 Members
3267 Online
110070 Solutions
New Discussion

Re: vg question

 
Filosofo
Regular Advisor

vg question

Hello,
I have a cluster with two nodes, Node-A and Node-B and PKG.
PKG active vg vg_try in exclusive mode (vgchange -a e)
Suppose that I want activate frome server Server-C the vg_try vg in read only,I must execute this command: vgchange -a r /dev/vg_try ?
PKG crash on Node-A what do happen to vg_try?
PKG deactivate vg_try from Node-A and reactivate this vg on Node-B in exclusive mode...and Server-C continue to see the vg in read only access without reexecute vgchange -a r after PKG switch?

I hope that I must plan.....I hope...

Filo
Sistem engeneer expert
9 REPLIES 9
Bryan D. Quinn
Respected Contributor

Re: vg question

Hi Filo,

I am not sure that this is going to answer your question, but this is what we do. We have a two node cluster, the primary being Node A and the failover being Node B. The pkg runs on Node A. When we add a disk to a VG, we activate it in read-only on Node B to make sure everything is fine with that VG on Node B. Then we deactivate it, here are the commands that we use:

Activate Read-Only (secondary node)
vgchange -a r vgXX

Deactivate (secondary node)
vgchange -a n vgXX

I hope this helps.

-Bryan
Geoff Wild
Honored Contributor

Re: vg question

Are your vg's on EMC?

If that's the case, then I would use BCV's - like a third mirror - but only seen by Host C - all you do then is splt the BCV's and mount them on Host C....

Is Host C in the cluster? If not then you can't do it with a simple vgchange -a r...

Rgds...Geoff
Proverbs 3:5,6 Trust in the Lord with all your heart and lean not on your own understanding; in all your ways acknowledge him, and he will make all your paths straight.
A. Clay Stephenson
Acclaimed Contributor

Re: vg question

You plan is really doomed. I am assuming that the shared VG uses block/io (e.g. filesystems or non-raw i/o).

Even in read-only mode, Server-C will not see consistant data. The problem is buffer cache on the active nodes. Server C doesn't have a clue about what is happening in Server A's buffer cache.

The only way that servers can share filesystem data (or other block/io) is for ALL nodes to be read-only.
If it ain't broke, I can fix that.
Filosofo
Regular Advisor

Re: vg question

the server-C is not in the cluster...why I don't execute vgchange -a r?
Sistem engeneer expert
Geoff Wild
Honored Contributor

Re: vg question

Why - cause MCSG will detect the vgchange from Server C as something like a disk failure - that's why the package fail's over....

You can only do something like this if you have a third mirror - like BCV's on an EMC frame....


RGds...Geoff
Proverbs 3:5,6 Trust in the Lord with all your heart and lean not on your own understanding; in all your ways acknowledge him, and he will make all your paths straight.
Filosofo
Regular Advisor

Re: vg question

Thanks for Helps...
But....what happens if pkg switch...Can PKG activate vg_try in exclusive mode on Node-B without deactivate the read only activation on ServerC?
For example If I stop PKG Can I see from Server-C, the server that have activate vg_try in read only mode before the stop of the PKG, vg_try in read only mode?
And what happans if I try to start PKG on Node-B without deactivate vg_try on Server-C?
....Another doubt...what are the difference between exclusive mode and read/write mode?

Thanks for yours helps....


Filo
Sistem engeneer expert
Filosofo
Regular Advisor

Re: vg question

OK I retry....
I have logical volume lv_try with FileSystem mount under /try....
I have activate, on Server A, vg_try with vgchange -a e.
On server-C I activate vg_try with vgchange -a r optinon.
Now I mount FS on Server-C under /try ....Now Ican read the files under /try also from Server-C.
Now I want activate vg on Server-B.
1 vgchange -a n on server A
2 vgchange -a e on server B

I mount FS always under /try.

Now I can read also from Server-C the file under Try?
Or whan I deactivate VG on server A I have also deactivate vg for server C

Please Help me is it important

Filo
Sistem engeneer expert
Michael Steele_2
Honored Contributor

Re: vg question

Please note the package.cntl file and the VGCHANGE variable.

#VGCHANGE="vgchange -a y"
VGCHANGE="vgchange -a r"
VG[0]...
LV[0]..., etc.

Please note the package.conf file and this variable: NODE_NAME [ node 1]
NODE_NAME [ node 2 ]

Order the nodes beginning with the primary node followed by the alternate for failing over to and in this way you designate how the package is to be handled.

Don't forget to mark the vg before starting the package:

vgchange -c y vg## (* must be run first *)

vgchange -a n vg##
Support Fatherhood - Stop Family Law
Michael Steele_2
Honored Contributor

Re: vg question

Also, for the 'three-server mutual takeover', you need to be using the -q quorum option. I.e.,

VGCHANGE="vgchange -a e -q n"

Try this link:

http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B5140-90011/B5140-90011_top.html&con=/hpux/onlinedocs/B5140-90011/00/00/13-con.html&toc=/hpux/onlinedocs/B5140-90011/00/00/13-toc.html&searchterms=cntl%7cvgchange&queryid=20030529-110510
Support Fatherhood - Stop Family Law