cancel
Showing results for 
Search instead for 
Did you mean: 
jest4kicks
Level 11

Configuring Cisco UCS Monitoring

Hey all, I'm in the process of adding our new (and first) UCS systems into Solarwinds, and there are a few questions coming up.

A little searching produced various bits of info, but much of it was from 2013 and prior.  Other threads present very limited sections of the process, so I thought we could get a thread going that covers the whole thing.

(Obligatory reference to an old KB on the subject: http://www.solarwinds.com/documentation/orion/docs/settingupciscoucs.pdf)

We're starting with 3 UCS chassis, and two fabric interconnects.  Solarwinds NPM is running version 11.5.2.

Configuration steps:

1. UCSM has been configured for LDAP authentication with our corporate AD domain.  Connectivity to UCSM has been configured over HTTPS, SNMP has been configured with a secure community string (fun fact, UCSM didn't like the "@" character in its community string), and a domain service account has been established with global read permissions to UCSM.

2. IP addresses for each of the fabric interconnects have been configured and noted, and the virtual IP for UCSM has also been configured and noted.  For the purpose of this example, we'll call them:

- Fab1 : x.x.x.1

- Fab2 : x.x.x.2

- UCSM : x.x.x.3

First Issue (Possibly Answered)  How are the nodes supposed to be added into Solarwinds?  Most instructions refer to adding the primary FI running UCSM first, configured to poll for UCSM information.  But if you do that, and UCSM switches to the passive FI, won't you lose UCSM monitoring?  Alternatively, I thought I could add UCSM by its virtual IP, but doing that still populates the resources from the FI its currently associated with.

Answer (Maybe): 3. I decided to add each FI as a node without UCSM polling, and then I added the Mgmt VIP with UCSM polling.  I configured each FI to poll all interfaces, and I configured the Mgmt node to only poll non-interface elements.  The idea here was for the Mgmt node to focus on UCS polling and let the FI nodes handle interface issues.  Polling some of the stuff at the top of the list might be redundant, but I wasn't sure if any of it was necessary for UCS blade info.

pastedImage_1.png

I'd recommend a service account for the UCS Manager Credentials section, and it will obviously need at least read access to UCSM.  When entering domain credentials, you have to specify the domain, but in the format that the domain is linked in UCSM (refer to the LDAP config in UCSM), AND with "ucs-" in front of it.

So for example, if your active directory domain is mydomain.corporate.realm, and you've added this into UCSM and called it just "ADmydomain", the credential has to be entered as "ucs-ADmydomain\mysvcaccount".

UCScreds.PNG

Fun fact, I had issues with the service account at first, so I added the Mgmt node without UCS polling initially, and then added the UCS polling once I resolved the service account problem.  (Ended up having to launch UCSM as the service account one time to register the account in UCSM.)  Doing this, the Mgmt node never successfully displayed UCS blade info.  I deleted the Mgmt node, then re-added it and included the UCS polling as part of the initial node creation.  Worked fine that time.

4. Once both FIs are added, you should be able to view the node that UCSM is active on and see the UCS Overview page element (ours showed up in the Network tab by default).

Second Issue: UCS Overview displays the status of each FI, and the names of the enclosures and blades, but the status of the blades doesn't populate.

Capture.PNG

I'll update this post with the additional steps and more pretty screencaps as we go.

Edit 1: 1/21/2016: Added a best-guess answer to the question of how to add the FIs and UCSM.  Raised the second issue of blades not populating.

Edit 2: 8/8/2016: It's been a while, but I'm giving this another go.  I think I found a better way of adding the FI's and the Mgmt VIP, so that info has been added.

Tags (1)
22 Replies
nollis
Level 8

Re: Configuring Cisco UCS Monitoring

Welcome to my world...I got our stuff at least added into NPM, the blades and whatnot do not populate though.  We do not have enough licenses to add EVERYTHING under both FI's and the VIP...so not sure if that is the problem.

To answer your question, I added the FIs in just SNMP and not UCS manager.  With the VIP, I added as UCS manager...that was what I gathered out of the other threads I found.

Good luck...if you blades populate let me know!!

jest4kicks
Level 11

Re: Configuring Cisco UCS Monitoring

Yeah I ended up doing the same thing, just to get something in there.  I'm not thrilled with it though, as I expect this will cause problems when UCSM ownership moves between FIs. 

Some good news: Solarwinds did get the names of the UCS blades.   The bad news is that I can't seem to get the status of the blades.  I thought I read that you have to add the management IPs for each blade as separate nodes in NPM.  I tried this, but there was no changes in the UCS Overview, and no obvious way to link the new node to the UCSM FI.  I'll add this as the next configuration step challenge in the first post.

Capture.PNG

0 Kudos
aaswi
Level 12

Re: Configuring Cisco UCS Monitoring

While I'd be the first to admit, Orion monitoring capabilities for UCS are limited.  There are several undp's floating around on thwack that can provide more specialized information. These have to be imported via the undp utility.

Chassis:

Cisco UCS Chassis Slot Information.UnDP

Cisco UCS Chassis Inventory with Status.UnDP

There are also some others floating around on here also I believe.

0 Kudos
designerfx
Level 16

Re: Configuring Cisco UCS Monitoring

jbiggleyadatole​  aLTeReGo have you guys worked with UCS monitoring outside of Universal device pollers?

0 Kudos
jbiggley
Level 14

Re: Configuring Cisco UCS Monitoring

I don't, sorry.  UCS gives me the heebee jeebees

designerfx
Level 16

Re: Configuring Cisco UCS Monitoring

I just thought something. After looking at Cisco UCS Active Hardware Alerts in Solarwinds​ and checking the UNDP, have you looked through what's available on the MIB doing an SNMPWalk starting around 1.3.6.1.4.1.9.9.719.1.1.1.1 or 1.3.6.1.4.1.9.9.719?  Also reference: MIB Reference for Cisco UCS Manager - Accessing Cisco UCS MIB Files [Cisco UCS Manager] - Cisco​  and/or MIB Reference for Cisco UCS Manager - Purpose of the Cisco UCS MIBs [Cisco UCS Manager] - Cisco  where it references monitoring the fabric interconnect? Or also "If your NMS cannot get requested information from a fabric interconnect or Cisco UCS Manager, then the MIB that allowsthatspecific data collection might be missing.". Theoretical here. Also Cisco UCS Monitoring Resource Handbook | Cisco Communities

It might (or might not) be too much effort to create a new UNDP or verify if this existing one is actually targeting the right part of the MIB in the first place, jest4kicks​ .? I've been looking into this for fun - however I would check if the existing MIB is correct? Could the UNDP be interpreting the data incorrectly?

0 Kudos
adatole
Level 17

Re: Configuring Cisco UCS Monitoring

A long, long time ago in a galaxy far, far away.

Seriously, haven't dug into the UCS side of the house in a looooong time. Maybe cobrien​ has some insight.

Leon Adato | Head Geek
------
"Measure what is measurable,
and make measurable what is not so." - Gallileo

0 Kudos
muckman
Level 11

Re: Configuring Cisco UCS Monitoring

Fun fact - UCS monitoring is not compatible with UCS-C240 servers. Hardware health will not be polled.

jest4kicks
Level 11

Re: Configuring Cisco UCS Monitoring

Well the good news for us is that we're primarily using B series blades. 

I had to take some time off from working on this, but I'll probably start cracking at it again soon.  Still looking for any advice from folks who have found their way through the muck.