NODE_TIMEOUT value?

[dictum9]

The NODE_TIMEOUT value is set to 6,000,000 in the /etc/cmcluster/ directory.

Does this mean it's set to 6 seconds?

I need it set to 8 seconds, how do I change it, and what needs to be done for the change to take effect after modifying the file?

[Pupil_1 · Accepted solution]

8 sec means 8000000 micro seconds.

[Pupil_1]

changes to the cluster.ascii will need a cmapplyconf and copy the binary to the other node to take effect !!
This will need a cluster service restart.

[melvyn burnard]

The NODE_TIMEOUT and HERTBEAT_INTERVAL are all in microseconds, so yes this is 6 seconds
To make the change, edit your ascii file and change the value, then halt the cluster, and use cmapplyconf to reapply the binary with the new value.

[dictum9]

Thanks, it's exactly what I was looking for.

[Steven E. Protter]

Shalom etc,

If you are planning on changing this value, caution is urged.

Most SG guides do have this figure increased a bit from the default to avoid unneeded TOC and boots of nodes.

But increasing it bears risks that your cluster response to real problems will be increased as well.

SEP

[Jeff Schussele]

This value should always be determined by 2 factors:

1) The avg latency of the network you employ as the heartbeat path - too high & you'll get unnecessary TOCs as SEP pointed out. We *always* use private nets with our own switches - you should too. We aaaalllll know that our public networks are light-speed & bullet-proof....right?????

2) The "urgency" your enterprise needs to get that pkg over to the adoptive node (notice how I've changed the term from failover - that latter word is now a no-no for us &~() )

In the end 8 seconds *should* be fine for 99% of clusters. I never have had to change it. And I don't see 6 as an issue either. Jeeeez if you can't get a low-level stack packet to your neighbor in 6 seconds you have bigger issues.....

My $0.02,
Jeff

[dictum9]

This makes it most clear. Thank you.