Spectrum able to Ignore High memory utilisation?

[Frank Tonjes]

Hi All,

We have some Cisco devices which by default use almost 100% for some of the memory in the device. This is normal operation, but we get a Major alarm for this. Is there a way to tell spectrum to ignore only these memory types (lsmpi_io)? I can see 2 memory entries in the Performance, but would like to only 'suppress' one of them. Any ideas on how you would solve this please?

Regards,

Frank

[Anon Anon]

Hi Frank,

What were the memory types available on these devices? Can you upload a screenshot? Do we have any specific objects in the Mib to differetiate these memories?

kalyan

[Frank Tonjes]

Hi Kaylan,

The Device is a Cisco ASR 1006:

[user@server vnmsh]$ ./show attributes mh=0x30070d

0x12aba NRM_DeviceMemoryUtilization_Threshold 85
0x12abc NRM_DeviceMemoryUtilization_Reset 70

0x12ac6 NRM_DeviceMemoryUtilization 1 69
0x12ac6 NRM_DeviceMemoryUtilization 14 100

0x12e3b NRM_DeviceMemoryUtilizationNames 1 Memory: Processor
0x12e3b NRM_DeviceMemoryUtilizationNames 14 Memory: lsmpi_io

Its the bottom one I'm trying to change. I see the thresholds are 'global' to this device. I was hoping to 'ignore' the lsmpi_io memory and still alarm for the Processor one.

I think this is one of the ESP modules and it's common for it to use all the memory.

I was thinking you could 'remove' the attributes from the table, but not sure what affect that would have regarding rediscovering it, etc.

Regards,

Frank

[Anon Anon]

Hi Frank,

You can create event condition to stop alerting on lsmpi_io pool or lsmpi buffers.

You can achieve this as below:
If the event contains the string like " Memory: lsmpi_io ", it should not alert and if it doesn't contain the "Memory: lsmpi_io", it should alarm.

For this we need two custom events: One for generating the event on "Memory: lsmpi_io" instance and one for other instances.

For Example: 0xfff00001 for generating event on Memory: lsmpi_io instance and there will be no alarm for this.
0xfff00002 for generating event on other instances and there will be an alarm. Select the probable cause code of 0x00010f06 to generate the alarm with the title "HIGH MEMORY UTILIZATION"

Then you need to create an event condition on 0x00010f06 to generate the events based on the value.


Hope this helps.


Rajashekar

[Frank Tonjes]

Hi Rajashekar,

Thanks, I was just looking at:

High Memory Utilization. The utilization of {C 1}% for memory instance {C 3} named '{C 5}' has exceeded the {C 0}% threshold on model {m} for more than the acceptable time period.

And if I can filter on '{C 5}' which it seems is possible by your mail then this can work.

Thanks guys.

Regards,

Frank

UPDATE:

I was able to create a Condition:

If {v 5} != {S Memory: lsmpi_io } generate event 0x10f06 and copy all variables. It seems to be working fine.

[VYNITA Pretorius]

We are on spectrum 9.3

I know this subject is old but .... I would like to ask if the event rule was added to which event was it added to

If {v 5} != {S Memory: lsmpi_io } generate event 0x10f06 and copy all variables. It seems to be working fine 

e.g.  SO the event rule (above) put on 10f06 then generated 10f06, can this be done?

Thanks Vee 

[Josh Bovee]

OK.  So this is fantastic information and I'll be applying it in our environment soon, but is there any chance this will get rolled out to Spectrum natively?  ASRs are very prevalent and as of 10.1.1 they're still being picked up as having high memory utilization out of the box.