An odd thing with volume groups and node timeout

[Simeon Harwood]

Afternoon all,
Has anyone seen this before?

My cluster setup is as follows:-
2 x D270
shared storage via scsi
HP-UX 11i
mc/sg v11.16

I wanted to change the NODE_TIMEOUT value to 30 seconds so that I can simulate network disruption during a switch upgrade.
1) I took the cluster down
2) edited the NODE_TIMEOUT to 30000000
3) cmapplyconf -C cluster.ascii
4) cmruncl
None of my packages started!!!!!
5) rolled my changes back again to original values of NODE_TIMEOUT 5000000
6) cmapplyconf -C cluster.ascii
7) cmruncl
Again, none of my packages started.
After investigation, I found that the volume groups had lost their cluster awarness, so I had to do a "vgahnge -c y vgxxxx" on all volume groups, then all was OK again.

So I repeated the test, and the same thing happened again!!!!
While I had my new value set to 30 seconds. I tried to do a "vgahnge -c y vgxxx" but was told that the volume group was not in lvmtab. So then I tried to import the mapfile again, and was told that the volume group was already in lvmtab!!!!!!!!

So now I'm confused. Is this a bug, or has anyone seen this before and I've missed a step out???

any help appreciated.

Ta,
Sime.

[Wayne Patton_1 · Accepted solution]

Are the volume groups in question listed at the bottom of the cluster.ascii? ie: VOLUME_GROUP /dev/vgXX?

If you do "strings /etc/lvmtab" you can see what vg's are in the lvmtab. I assume your vg's are all in there now.

wayne

[Simeon Harwood]

Good man Wayne,

You've just made me realise what I've done.
I was using the original ascii file and not a new created from the running cluster. So the volume groups are not in the ascii file.

Good man,

Thanks for pointing out my mistake!!!!

Sime.