Idea Details

customizable discovery configuration on policy basis

Last activity 06-13-2019 09:20 AM
Andre Heller's profile image
01-26-2015 04:22 AM

What is the use case of the new feature?

----------------------------------------

Customizing discover policies regarding get bulk options to successfully discover special devices.

 

Describe the new feature in detail

----------------------------------

PM 2.0 is using get-bulk requests for the model discovery.

However, some devices do not deliver all requested OIDs in the SNMP response. E.g. if an SNMP get-bulk request contains a high count of max-repetitions the router may not respond with the total count. This is correct according to RFC3416 (for further details have a look at: http://kb.juniper.net/InfoCenter/index?page=content&id=KB7781)

As a result the device is not discovered and modeled correctly.

As you can see, hard coded values as in the current design (e.g. the max-repetitions value) are not usable for every case and the discovery configuration does not provide the required support for heterogeneous environments.

 

Describe how you envision this new feature being implemented.

-------------------------------------------------------------

Discovery configuration should be more customizable on a per policy basis.

 

What business problem will be solved by adding this new feature?

----------------------------------------------------------------

Devices with non optimal agent configuration or a different design could be discovered without additional configuration on agent side.

In most environments it's much more complex to re-configure agents than to configure central management console (e.g. do firmware upgrade or getting access permission for reconfiguration). Therefore an enterprise solution like PM 2.0 should provide possibilities to customize the discover settings individually.


Comments

08-07-2017 01:53 PM

Hi Andre_H,

 

I discussed this briefly with the engineering team.

 

They explained that by default, CAPM uses a max-repetition value of 10 when performing a getBulk.  If the request returns tooBig, we adjust the max-repetition to half it's previous value and try again, until it reaches 1.  Therefore, we wouldn't expect any issues here.

 

I know this is an old idea, so it may be an issue that's already been addressed, since the idea was opened.

 

If anyone is experiencing this problem where devices are not being discovered properly on a supported version of CAPM, please open a Support ticket so we can troubleshoot.

 

Thanks,

Matt