Start a Conversation

Unsolved

This post is more than 5 years old

1470

November 9th, 2009 11:00

Use Dynamic Model for adding "hosts xyz" ?












:-)















49 Posts

November 9th, 2009 11:00

Hi Fred, edit: and Berkay

thanks for the fast response.

I have choosen the "blade" stuff only to make it understand better by others. Maybe it was not my best idea tonight :-). Let me try to explain with the correct syntax.

First of all, it's all software related, MIBs cover everything needed.
We have several servers running a certain software (DNCC service), one service per server.
Each DNCC contains several InrouteGroups.
Each InrouteGroup contains several Inroutes.

"Topology":
When walking the MIB, I can see:
- which InrouteGroups are configured on the DNCC
- which Inroutes are configured onthe appropriate InrouteGroup
So we can see, what is related to which server/DNCC. This should answer the topology question I guess.

I created the DNCC class from the host-template, maybe its better to use the switch/router as its "interfaces" can reflect my groups ? And also use switch/router for the InrouteGroups to have their "interfaces" as inroutes ?
Maybe thats a way to use already implemented dependencies for discovering ?

I never worked with those part of smarts in depth before, just scripted around what was already there, so its quite confusing.

edit:
I have no possibility to access an Cisco WAN Manager (ATM/FR) installation.

Just read both posts again. It seems the question itself is already answered and Dynamic Model can't be used to reach my goal. Or even, can't be used for all I have to do.
Its partly a discovery/instrumentation thing, where I have to deal with all those ASL-scripts and job for Dynamic Model to add certain attributes to a class, which values are also polled by the Dynamic Model part ?

Which documentation is recommended to get familiar with all those discovery stuff ?

Thanks for patience

Rgds Christian

Message was edited by:
choefing

36 Posts

November 9th, 2009 11:00

Hi Christian,

I wrote a long email and got stuck at the end, and starting over :)

What you're describing sounds similar to monitoring a network via element manager rather than directly dealing with the individuals directly.

I think Smarts domain for Cisco WAN Manager (ATM/FR) works as you describe, meaning Smarts polls the Cisco element manager, discovers the devices and monitors all via SNMP without any direct communication with the devices. So if you can access to that product, you can learn a lot from it.

Bottomline is I think this is rather complicated undertaking. Don't think Smarts domain handles instrumentation by proxy easily. Your challenge is not really modeling part, but getting the discovery (not so difficult) and the instrumentation (difficult) sorted out.

Hope this helps,

Berkay

November 9th, 2009 11:00

Hi Christian,

"instrumented SNMP" is for monitoring, not for discovery. So, yes, from what I understand from your description, you will have to add a custom discovery process (not so hard, depending on the MIBs complexity) in order to discover things and create the right relationships between them.
First question : do you have the topology information (which blade belongs to which blade server and the HW stuff - index, ...) in the MIB itself ? In the blade server MIB ? In the blade MIB (if any) ? If this information is available in the blade server, you will have to do a custom discovery of this device and create/index the discovered blade server.

HTH,

--Fred


== Monitor your Smarts environment using APG ReportPack for Smarts health ==
Frederic Meunier
Solutions Watch4Net Inc
APG & Smarts InCharge integration
http://www.watch4net.com

November 9th, 2009 13:00

Hi Christian,

as I don't know "DNCC", I won't be able you a lot here. My first comment is : are you sure the Switch/Router attributes/relationships/behavior or the interfaces ones are similar with InRouteGroups and InRoutes concepts ? You are trying to fit the existing ICIM to something else, this may not be the right solution.
Anyway, if those concepts are similar, what you can do : use the oid2type mechanism to discover instances of your classes as you did (DNCC, InRouteGroups, InRoutes). Then, have a specific custom post discovery process for DNCC instances, which will discover relationships between DNCC and InRouteGroups / InRoutes and make the association between corresponding (and already existing) instances of the discovered objects. If you discover the DNCCC before the other objects, the objects won't be related at this step, but at the next discovery process, then, as all the objects should exist, the post process script will have all the data to build the relationship.

--Fred


== Monitor your Smarts environment using APG ReportPack for Smarts health ==
Frederic Meunier
Solutions Watch4Net Inc
APG & Smarts InCharge integration
http://www.watch4net.com

49 Posts

November 9th, 2009 15:00

Fred,

this DNCC stuff is just a piece of proprietary software managing/controlling certain hardware where we need to monitor the status (something like interface up/down or error rates). All we need can be read from the snmp-agent running on the server.

With SAM/IPM 6.0 we alrady got an "custom poller" from EMC, but that was before I joined the company, so I have no really glue whats hardcoded in there.
As there were some changes and we were in need to gather more information they told, that we should migrate to 7.0 and use dynamic model to build what meets our needs.
To keep the business critical monitoring working they adapted the old custom poller and build again somethnig hardcoded but this is still missing the new values.

I guess my fault was to think, that I can use dynamic model only to reference everything needed to discover, build dependencies and so on..

Let me give an example of what exactly is needed.

Each DNCC has 6 InrouteGroups
Each InrouteGroup has 70 Inroutes.
(its the same on every DNCC)

The difference is the number of InrouteGroups/Inroutes which are active. You can compare this with a port on a switch administratively set to up or down.

The old solution on 6.0 added all Inroutes and all InrouteGroups, which did not react to the smarts-setting "isManaged". I guess it was only not checked as the MIB was not fully covering this years ago. So we got a lot of false negatives.
An Inroute, which is not configured on the DNCC, should not be shown as down in Smarts.
I tried to work around this issue setting up an observer using perl waiting for Inroutes to go down and check in a hardcoded array if this is an configured Inroute or not.
This works as long as no changes were made inthe configuration, what is quite unusual in our setup.
So the idea was born to check for the administrative status before an escalation is triggered.

I hope its not completely confusing you. I know that it's a quite unusual thing to setup.

As far as I read now and additionally what you and Berkay wrote... I have to work on 2 parts.
The part where a notification depends on the administrative status of an Inroute has to be done with an additional attribute for the Inroute, gahtered with instrumentation via SNMP.
The discovery part has nothing to do with dynamic model itself and I have to deal with that as well as the hardcoded poller does not work correctly anymore due to API changes in the DNCC.

Thought about the "switch" question. I actually don't know what class meets the 3 parts the most so maybe I better start off with annearly empty one and know afterwards more about the internal things going on.

When I invoke a dmctl -s SAMSERVER getClasses, I get hundreds of different classes. What would you suggest to do to find the most matching one at least for the DNCC, which is also a normal piece of hardware, a server. For now I use a "copy" of host (ICIM_something? .. can't say for sure at the moment)

I already used the oid2type to discover the DNCC itself. That works perfectly.
I#m sure that I also could do it with every single InrouteGroup and Inroute.. but..
What I'm missing is the "how do i get the list of inroutegroups (automatically) it should discover afterwards" as we don't want to seed any group and any inroute, even not in a bunch, because it worked with the hardcoded customer poller in 6.0. with just discovering the DNCC itself.

It makes sense, that I first have to discover everything, before I can setup relationships between them.

Think I have to read a lot within the next days.
As for now I will set this to answered, as both told me, that I can't do this with Dynamic Model as standalone solution.

Thanks so far for the explanation. If you have further information or hints regarding this, feel free to just post below.

rgds
Christian

18 Posts

November 9th, 2009 17:00

Hi Christian,

This is Karl from Hughes in Germantown. I'm familiar with your custom configuration that was done many years ago to monitor the different Hughes servers and their enterprise MIBs.

To be honest with you, what was done and what you are trying to re-create/update is very difficult to do with SMARTS AM/PM. It is like a square peg and a round hole. We have built some very extensive monitoring tools that monitor all of the Hughes components and feed that analysis into SMARTS via traps.

SMARTS is then used for Infrastructure, connectivity, host resources, and event correlation.

Please feel free to contact me if you would like more information.

Karl Fosburg

49 Posts

November 9th, 2009 22:00

Good morning/evening Karl,

thanks for the response.

Glad to hear, that I am not the only person struggling with that, even, if the best solution would not be, what we expected when buying the dynamic model license.

I'll drop you a line.

Thx & rgds
Christian
No Events found!

Top