Service Profile Failover

[JoeKeegan3]

I don't think this feature exists, yet, but wanted to see if anyone has a way to do it or knows if it's coming.

Is there a way to have the UCS Manager "Failover" a service profile if the current blade that the service profile is running on dies. Something like this:

- Blade Dies (CPU failure, yanked out, etc)

- UCS manager detects that the blade has died

- The service profile is disassociated with that blade

- The service profile is associated with a new blade

- The new blade is powered on

Obviously there would be caveats, must not use local disk, etc.

Is there any way to do this today?

Thanks,

Joe

[Manish Tandon · Accepted solution]

Joe

Integrating SP failover as part of UCSM is something which has been thought about and is being looked into.

It has not been concept committed hence no timeline associated with it.

Currently you can do it via a system external to UCS though.

If you look at the developer community for UCS at http://developer.cisco.com/web/unifiedcomputing/resources/-/document_library/view/2144116

you will find failover ServiceProfile.zip which does pretty much what you mentioned. As its a developer forum you can modify/contribute to it too.

No doubt this functionality will be good to have in UCSM instead of having an external system poll it like in this case but still is a very useful script to have.

Hope it helps.

Thanks

--Manish

[JoeKeegan3]

Thank you for the information.

[dani_bosch]

Hello all,

It's been a year from this question, and feature is not there yet, right? Any inputs on whether Cisco does or doesn't have the intention of ever implementing it?

Thanks a lot,

[Manish Tandon]

Dani

As I communicated to you in the unicast thread, this feature of full blown blade/SP failover has not made it in the list of upcoming features.

A matter of feature prioritization is what it eventually boils down to.

I will not go so far as to say it will be never, but its not in the immediate timeframe.

Thanks

--Manish

[colin.lynch]

Hi Dani

I think the closest you can get to this nativley in UCS today is to associate the SP to a server pool, if that pool has an available blade in it, the SP will dynamically move to another blade if the current blade is put in a decommisioned state.

I would think that adding some sort of proactive polling and leaving UCS to dynamically move the SP, would come with another set of issues, i.e. would you want the system to move a profile off a server if that server was reporting memory errors for example, almost certainly not as moving the SP does currently involve several minutes of downtime for that SP.

More of an argument if the blade died completley but less likley than a degraded situation like the above, inwhich case you would probably want your SP to stay up until your first available maintaince window and then move it.

But Manish is the man on this, and as he says a few things to address first, (renaming SP's for a start :-)

Regards

Colin

[dani_bosch]

How do HP and IBM handle this with Virtual Connect and the IBM counterpart? Do their solutions work well? If the answer is YES: ¿Why isn't Cisco doing something similar?

Many thanks,