reasons to bring down cluster

[khilari]

Hi Guys, i am looking for situations where i have to bring down my cluster environment to perform necessary changes.
Sometime we can make changes online but which situations require cluster shutdown.
Thanks

[Stephen Doud]

I recently submitted a document for the customer Knowledge Database that discusses this very topic. It is titled, "HPUX Serviceguard - permissible online cluster and package modifications". It summarizes all cluster and package parameters and other cluster admin tasks on currently supported versions ie A.11.16-A.11.19. Unfortunately, it has not made it through the editing function yet, so please email me to request a copy.

[Steven E. Protter]

Shalom,

1) Changes to shared storage.
2) Updates that impact Service Guard.
3) Loss of hardware risks data corruption (precautionary, SG still runs but the risk is too high to remain in operation)

Neither is mandatory but the purpose of clustering is to avoid other down time issues.

SEP

[khilari]

Hi Stephen Doud, thanks for your post. What is your email where i can ask for the document.
Thanks

[Rita C Workman]

Not sure about your SG version, and this may not be complete...but...

Generally the must be halted if you want to change:

Timing Parameters
Cluster lock configurations
Changing IP address
Changing serial device files

That's all I can think of real quick....

Regards,
Rita

[Ron Irving]

Pardon me for jumping in...I'd like a look at that doc too!! I just inherited an SG environment here at a new job.

Thnx!!!

No pts for this please.

[Rita C Workman]

...and a couple obscure reasons:

Moving the data center
Testing to ensure your cluster will start up everything as it should properly.

/Rita

[Viktor Balogh]

The cluster must also be stopped if you want to reconfigure it, (eg. adding/removing a new node)

[Stephen Doud]

stephen.doud@hp.com

[Gordon Sjodin]

Cluster related patching requires downtime.

[Rita C Workman]

One correction - I don't believe the cluster has to be down to add or remove a node.
And I believe you can actually remove a node even if it (the node) isn't running at the time of removal.

/Rita

[James R. Ferguson]

Hi:

@ Stephen: You might want to have your email address reformatted by a moderator to make it less likey for bot harvests. :-)

Regards!

...JRF...

[Viktor Balogh]

here is a table about the cluster state during the different reconfiguration steps:

http://docs.hp.com/en/B3936-90144/ch07s04.html

Note that this is for the latest SG version which is A.11.19

[Chetan_5]

After going thru the link above, it seems like pretty much all maintenance can be done online, except for maybe changing the cluster name.

Chetan

[Stephen Doud]

Even for A.11.19, which permits the greatest breadth of online reconfiguration, not all maintenance can be done online safely.

I recommend a review of the edition of "Managing Serviceguard" table 7-2 that matches the version of Serviceguard in use, to find those items that should not be modified during package operations.

Managing Serviceguard (A.11.19 first reprint edition):
http://docs.hp.com/en/ha.html#Serviceguard
Table 7-2
http://docs.hp.com/en/B3936-90144/ch07s06.html