11.16 NETWORK_FAILURE_DETECTION

[Tonya Underwood]

I am preparing to upgrade many clusters to 11.16 from 11.15.

I see that NETWORK_FAILURE_DETECTION defaults to INONLY. I will be changing this to INONLY_OR_INOUT. Will it still be possible to perform a rolling upgrade? Or must I have cluster down to reapply?

Thanks,
Tonya Underwood

[Steven E. Protter]

Hi Tonya,

Don't know.

Did you happen to read this document?

http://docs.hp.com/en/B3936-90024/apes02.html

Its a rolling upgrade example.

SEP

[Tonya Underwood]

Yes... read this document as well as the Managing MC/SG for 11.16 (12th edition)... nothing states whether I must re-apply the cluster if I change this variable...

I may just have to do a test cluster and attempt without bringing cluster down...

[Steven E. Protter]

In that event I suggest you try a rolling upgrade in your test environment or your least important(if such a thing can be said) production environment.

Or perhaps call your local response center. They may be able to run a short experiment on someone elses hardware.

SEP

[Tonya Underwood]

Yes... I will try rolling upgrade first... if the new value for the variable does not change, I will bring the cluster down and reapply/restart...

[melvyn burnard]

This parameter is not an online reconfigureabke one I believe.
I will verfiy this, and also request the manual be updated to specify whether it is or not, as it is unclear in teh current version

[Tonya Underwood]

Thank you so much! I look forward to your response!

Tonya Underwood

[melvyn burnard · Accepted solution]

Well I can now say that I have tested modifying this parameter with a cluster up and running at 11.16, and it DOES work as an online change.
I am busy verifiying that this is a SUPPORTED online change.

[Tonya Underwood]

That is great news. Let me know what you find.

[Stephen Doud]

This whitepaper indicates that online reconfiguration change of the parameter IS supported:
"Serviceguard Network Manager: Inbound Failure Detection" ---> http://docs.hp.com/en/5568/SG_Network_wp_final.pdf
See page 4, first paragraph:
"The network failure detection method is configurable, so users can choose between the existing default method and the new enhanced method. Also, users will be able to apply the configuration online and make their choice without having to bring the cluster down. They can do this by modifying the cluster configuration file and setting the value for the NETWORK_FAILURE_DETECTION parameter."

[melvyn burnard]

Well this is indeed an on-line configurable parameter as listed in the White Paper that Stephen pointed out to you.
I have requested the Managing Serviceguard manual be updated, which will be sometime in the New Year. Unfortunately we have missed the deadline fo rthe next release of the manual.

[Tonya Underwood]

Thank you all.