Service Guard

[dan_156]

Question? Can you form a cluster consisting of 1 x V2200 running 11.0 and 1 x RP8400 running 11.i? Will the different OS be a issue? We are looking a porting our env over to a RP8400 and kicking the idea around of using service guard to do this.

dan@matlock.com
or
dan.matlock@imagesoft.fiserv.com

[John Poff]

Hi,

I'm pretty sure that if you run version 11.13 or later of MC/SG you can have different versions of HP-UX on the nodes.

JP

[Anil C. Sedha]

I don't believe it's possible to have a cluster between 11.0 and 11i

-Anil

[Jeff Schussele · Accepted solution]

Hi Dan,

Although it possible to run mixed OS versions (This is what happens in a rolling upgrade) it's certainly NOT recommended to run for an extended period this way.

But if all you're doing is using this to port the app over & you'll disable the cluster shortly afterwards - you can probably get away with it.

Rgds,
Jeff

[melvyn burnard]

Using a cluster of Mixed OS versions 11.00 and 11.11 is SUPPORTED, provided that all the nodes in the cluster are running the same version of ServiceGuard, which must be either 11.09, 11.13 or 11.14
One thing to be carefull of is if you have systems running the 32bit version alongside systems with 64bit HP-UX, but looking at your question, these will all be 64bit systems.

[dan_156]

Excellent feedback! This is only an idea of mine right now, but we need to move production off of an old V2200 to two RP8400s in a cluster. Instead of a HDware 'big bang' I thought of just failing over and then later dropping the V from the cluster. Now I just need to work out the different fiber/HDware pathing for the VG inport/export scenario. We are using EMC 8830s. Again thanks to all for the fast responses. Last service guard I did was 2 x RP7400 clones, same HDware and a slam dunk.

[monasingh_1]

SEvice guard will work and This is used to do the rolling upgrade. But remember that applications like oracle will not work, becasue they need recompilation once they go from 11 to 11i. So service guard will work, mount will happen but if some like oracle is involvedm, that wil fail and package will go down again depending on the configuration.

So, technically, it will work, but practically, will depend on your environment..

for porting things from 11 to 11i, if you can get some down time , then disable the application startup from MCSG and the just mount the respective directories and then do the recompilation as needed and then enable the application startup from SG.

I would still test the recompilation and validate before proceeding..

[dan_156]

Excellent feedback! We use Sybase and install/run the same version on 11.0 and 11i and have a warm fuzzy that it will work. My main concern now is trying to export/import the VGs from the V Class to the RP8400. Straight forward with maps on clone HDware/IO slots, but will make me think a little now with dissimilar HDware.

[monasingh_1]

vg export/import is not hardware dependent. All info regarding the VG and the required disk is written on disks so you should not have any problem.

If you want to have a comfort level then for test you can create a VG on V class and then use export and then try to export it on rp series. It should work as long as you can see the disk.

You will need to change the cluster lock disk path name in the cluster config because the cluster disk will most probably be at a different path on RP8400.

Or as an alternate you can use quoram server instead of cluster lock disk

[dan_156]

Excellent. I didn't think that far ahead on the VG Lock Disk path. Will I need duplicate entries do to LOCK_PV although same VG will have different HDware path, or two config files? (1 x for V and 1 x for RP)

FIRST_CLUSTER_LOCK_PV
/dev/dsk/c7t5d4

[melvyn burnard]

you need to know which is the pv and device file you ar etrying to use for the cluster lock diosc.
If all else fails, do your shared vgimports, and then use the cmquerycl comand to create a new ascii file, and SG will try to define your cluster lock disc for you.
You can then set it correcttly in your ascii file and reapply the binary, with eth CL VG activated.

[Brian M Rawlings]

Dan: I don't have anything to add, these guys have helped you with lots of good information. I am curious, however, if perhaps you are not familiar with the practice on this forum of assigning points to answers, both as recognition to those who have helped you, and to note particularly helpful answers (8-10 points cause the "magic rabbit" icon to appear with the question for future researchers).

If you'd like to familiarize yourself with the guidelines for this voluntary activity, check out:
http://us-support3.external.hp.com/estaff/bin/doc.pl/screen=estaffAssistance/distrib_redir=2+1046103149|*?Page=file0002#forassign

If you are just morally opposed to point systems, nevermind. I just thought that several of the excellent answers above deserve some point recognition, if you have the time.

Best Regards, --bmr

[dan_156]

Was not familiar with the points system, but am now. I will gladly give points. Everyone has been helpful.