« Back to Technical Discussions

how to simulate failures in UCSPE?

Combination View Flat View Tree View
Threads [ Previous | Next ]
 
Is there way we can simulate failures like (low voltage, low temp, low bandwidth stats..) in UCSPE?
 
Thanks

Hi Shiva

You may find the failure details in XML response
Search for node faultInst in response, If node exist that means there is some fault . Also check for the severity attribute, if its value is cleared that means it has been resolved

Thanks for response Dilip.

Do you have an example where I can inject a failure (low voltage, low temp..) in UCSPE through api?

Shiva

They are automatically injected by UCS in response as faultInst is a read only class and u cant send / inject faultInst class id or any of its attributes in to your XML API request

So, I can't simulate any failures which will alter 'Stats' in UCS emulator too?

Thanks

May be in some other way u can achieve or may be some workarounds could help but before going forward to a solution, a complete discussion of use case is required

Is there way we can simulate failures like (low voltage, low temp, low bandwidth stats..) in UCSPE?
 
Thanks

 
There is no way to simulate failures in UCSPE today, nor any way to affect values returned by the emulated sensor values of the Fabric Interconnect / Chassis / Blades, etc.
 
It is something that is on the roadmap for a future release, but not committed at this point as to a specific timeline.  When implemented, there will be an API to inject fault conditions within the emulator to trigger issues that will cause UCS Manager w/in UCSPE to raise faults.
 
Thanks,
Eric Williams

Thanks Eric for the response.

Do you know how did the cisco team tested those types (stats, perf counters) without any simulation or fake failures?

Basically, we need to collect perf and capacity counters of UCSs which vary every interval but in real devices we can't do it for every UCS instance (too much work load).

Thanks

Thanks Eric for the response.

Do you know how did the cisco team tested those types (stats, perf counters) without any simulation or fake failures?

Basically, we need to collect perf and capacity counters of UCSs which vary every interval but in real devices we can't do it for every UCS instance (too much work load).

Thanks

 
There are debug capabilities within each device (CIMC, Fabric Interconnect, blades, etc) where Cisco engineers can override the data to trigger faults and such.  That's a similar plan with the UCSPE fault injection API we plan to implement in the future.
 
Eric