Would not be alright to put both Dev and Prod servers onto the same Superdome?

[Hanry Zhou]

and why?

please let me a few concrete examples if it is not a good idea.

Thanks!

[Larry Klasmier]

Are you going to partition the server? if not you will be applying OS to both production and Dev at the same time, so you have no way to test prior to potentially affecting production.

[Torsten.]

If you don't use your development system as a backup - why not?

[Dennis Handly]

What type of problems are you worried about?

Developers slowing down production? Or crashing production?

Here is where Larry's partitioning will help isolate the two systems.

[Hanry Zhou]

I know I can put Dev and Prod into two different Npar, but I am worried about something may happen that could crash the entire supersome and therefore Development may bring down the production.

Will there be any situation like that ?

[g3jza]

one nPar is electrically isolated from the other so it is less probable to crash than for example having 2vPars on the same nPar.

[Duncan Edmonstone]

There are still a few SPoF (Single Points of Failure) in a Superdome 2, so there is still the outside possibility that something could take out the entire complex. I wouldn't be concerned about these issues particularly though as they are generally for components which have execptionally long MTBFs (e.g. the SD2 blade enclosures backplane which is just etched copper with no active componentry) - I'm simply making the point that these systems are not designed to be "fault tolerant" in the truest sense of the phrase.

That said, from an OS perspective, there is NOTHING that an OS in one nPar could do that would cause another nPar to crash - the only issues you can have are issues that effect the whole complex (and you'd get those even if you had seperate complexes!), or security issues (some nervous customers will simply never allow production and dev/test equipment on the same complex, beacuse it generally implies shared adminsitration/access which they might not allow)

HTH

Duncan

[點燃]

saludos Hanry,

I would not say it's completely a bad idea, still there are chances that you run into unexpected issues. Let me mention an issue I've faced in my environment. We had a test and prod npar on the same complex and had to shutdown the test npar for CPU replacement on one of the cell. After replacing the CPU, when the HP CE tried to boot the partition, it was not coming up. After trying many options (without connecting the faulty cell at all), he confirmed the backplane is faulty and to replace it we need the entire complex to be down. Remember the backplane is a SPOF in superdome as well. However this is a rare incident and we still don't know what caused it.