44 Replies Latest reply on Jun 22, 2017 12:35 PM by JesperVestergaard

    No, really, how DO you get UCS Blade Status?

    dmeiser

      A topic that much has been said about and people have gotten working, but I can't find how anybody fixed it!  Does anybody have a step-by-step they followed to fix this:

      Screenshot from 2017-02-24 16-22-57.png

        • Re: No, really, how DO you get UCS Blade Status?
          designerfx

          I believe jest4kicks had a thread working on that

          • Re: No, really, how DO you get UCS Blade Status?
            rschroeder

            Wow.  I expected / assumed that you'd get full information from UCS into NPM by creating the SNMP community read-only string on the UCS chassis and then discovering it with NPM.

             

            Then I expect more information would be available from the servers on the blade by configuring them for WMI and discovering them via WMI with NPM.

             

            Are you telling me this doesn't work?

              • Re: No, really, how DO you get UCS Blade Status?
                dmeiser

                I had assumed the same.  I had assumed that this was something I was doing wrong, too.

                  • Re: No, really, how DO you get UCS Blade Status?
                    rhidians

                    I would love better clarification on this..... But as far as my understanding goes (And it is limited) this a limitation of SNMP, not really Solarwinds.

                    I don't think SNMP was designed (ahem, 5 or 6 decades ago?) to include SNMP community within SNMP communities. Or multiple SNMP communities under one SNMP community and that is what UCS is.

                     

                    A similar situation is a box that houses multiple vm's (I forget the name of the firewall I'm thinking of) and each of those vm's has it's own SNMP mibs.

                    It would be nice if we could get an overriding MiB that communicates with the sub MiBs (definitely needs clarification). But again, I don't think SNMP was imagined to be used like that.

                     

                    I suppose an ESXi server is a good example, we can get the SNMP information from ESXi but we cannot get the ESXi SNMP to give use the SNMP status of the internal workings of the VM's running on it. In order to do that we need to use the VMware API.

                     

                    Like I say, I'd like to be clearer on this as it's some thing that comes up quite often for me.

                    Hopefully I am wrong and there is a FAR better answer, especially one that goes: It's done like this!

                • Re: No, really, how DO you get UCS Blade Status?
                  rschroeder

                  Have you checked out this process?  monitoring ucs

                   

                  If you don't find a satisfactory solution, I think it's time to call Solarwinds Technical Support and request assistance.

                  • Re: No, really, how DO you get UCS Blade Status?
                    rschroeder

                    It's disappointing if this documentation, which appears to show NPM displaying useful UCS discoveries, is in accurate:

                     

                    Orion NPM 10.1 Meet the Features – Cisco Unified Computing System (UCS) Support

                    • Re: No, really, how DO you get UCS Blade Status?
                      Vinay BY

                      Hi dmeiser

                       

                      You can add each blade as a separate node and status gets linked onto your UCS Overview Page.

                       

                      Go back to same page UCS Overview, left mouse click on blade-1 and add it to monitoring(/manage) on your Solarwinds. Then refresh your UCS Overview page and check if you can see the status of the blade (after you add it as a node to solarwinds).

                      Follow the same for rest of the blades.

                       

                      You might not like this, but yes it works.

                        • Re: No, really, how DO you get UCS Blade Status?
                          designerfx

                          This actually makes sense to me. I was under the impression this didn't work somehow from what jest4kicks and dmeiser were saying?

                            • Re: No, really, how DO you get UCS Blade Status?
                              jest4kicks

                              Yeah, still not working for us.  Our blade names are not actually click-able in the interface.  The blades themselves have already been added as single nodes in Solarwinds, but whatever mechanism is supposed to link the nodes to that interface doesn't appear to work.

                               

                              1 of 1 people found this helpful
                                • Re: No, really, how DO you get UCS Blade Status?
                                  mesverrum

                                  When you added the blades did you add them from the UCS manager's details screen or did you add them from a discovery/manage nodes?  Are you seeing the from the Solarwinds node associated with the manager and not from the addresses for the fabric a/b?

                                   

                                  If you are savvy in SQL you can pick apart what is going on in the relationships by looking at [dbo].[NPM_UCSManagers], [dbo].[NPM_UCSChassis], and [dbo].[NPM_UCSBlades]

                                    • Re: No, really, how DO you get UCS Blade Status?
                                      jest4kicks

                                      What do you mean by "add them from the UCS manager's details screen"?  I'm just looking at the UCS FI that we added as a node in solarwinds (technically the management name/IP, not a specific FI).  I don't see anywhere to add nodes here (as opposed to a vmware host, for example, which lists additional nodes and allows you to click and add them).

                                       

                                      We had to add all of our blades manually because of this.  I should point out that I also tried manually adding the blade CIMC IPs as nodes, manually, but they still didn't link with the UCS FI node.

                                  • Re: No, really, how DO you get UCS Blade Status?
                                    mesverrum

                                    I think where people's expectations get messed up here is that the IP solarwinds is polling under blade 1-8 is actually the blades' management IP within the UCS and is not integrated with the OS that is on the blade in any way so they don't see what they are expecting/hoping for.  It tells you that a blade is plugged in and answering pings and not much else.  You just click on blade 1, it tells you this node isnt being managed in solarwinds and asks you if you want to add it and you select ICMP only and then it is done. 

                                     

                                    I think a lot of people expect that there is some way that Solarwinds is going to map from the Chassis to the blade to the OS instances that lives in that blade. Based on the times I have had to dive into this for clients I recall that using custom SNMP pollers from the UCS manager you can poll a lot of info about the hardware capabilities of the blades and their overall up/down status, but actual resource utilization outside the fabric interconnects is sparse.  It doesn't tell you anything about the OS installed on that blade or how much CPU it is using or anything like that.  I've rigged it up before where I could click on a UCS chassis and drill from there down into the actual solarwinds/vman ESX host objects by butchering the database and manually changing things around, but there would be no way I've come across for Solarwinds to know how to correlate that info based on what UCS/ESX tells it, you have to manually correlate it all.

                                      • Re: No, really, how DO you get UCS Blade Status?
                                        dmeiser

                                        mesverrum wrote:

                                         

                                        I think where people's expectations get messed up here is that the IP solarwinds is polling under blade 1-8 is actually the blades' management IP within the UCS and is not integrated with the OS that is on the blade in any way so they don't see what they are expecting/hoping for.  It tells you that a blade is plugged in and answering pings and not much else.  You just click on blade 1, it tells you this node isnt being managed in solarwinds and asks you if you want to add it and you select ICMP only and then it is done. 

                                         

                                        I think a lot of people expect that there is some way that Solarwinds is going to map from the Chassis to the blade to the OS instances that lives in that blade. Based on the times I have had to dive into this for clients I recall that using custom SNMP pollers from the UCS manager you can poll a lot of info about the hardware capabilities of the blades and their overall up/down status, but actual resource utilization outside the fabric interconnects is sparse.  It doesn't tell you anything about the OS installed on that blade or how much CPU it is using or anything like that.  I've rigged it up before where I could click on a UCS chassis and drill from there down into the actual solarwinds/vman ESX host objects by butchering the database and manually changing things around, but there would be no way I've come across for Solarwinds to know how to correlate that info based on what UCS/ESX tells it, you have to manually correlate it all.

                                        This, I think, is the problem that I'm having.  I'm expecting some sort of blade health status - not OS level information, but hardware health such as disk and whatnot.  From what it sounds like, that expectation is too much and I need to expect ICMP/ping status.  Is that accurate?

                                  • Re: No, really, how DO you get UCS Blade Status?
                                    lfransen

                                    So here is something interesting I found out. When you add your UCS Manager node you add it as ICMP only, then select 'Poll for UCS' and add the appropriate credentials. Then I got to see the chassis/blade/switch info. From there you have to add each component separately as individual notes. Now you should see everything and its relationship when you go back to looking at the UCS Manager node.

                                     

                                    If you add the node as SNMP, it seems like you can poll all the info, but all my Chassis/blade/switch info disappeared. Not sure if this is the correct behavior but that is how it worked for me.

                                      • Re: No, really, how DO you get UCS Blade Status?
                                        dmeiser

                                        This is exactly where I get confused.  I get the Fabric Interconnect info, but I am never able to populate the blade status.

                                        For those who get blade status:  Are you monitoring the blade IP address (eg - the KVM access address) or the IP address of the Operating System running on the Blade?

                                          • Re: No, really, how DO you get UCS Blade Status?
                                            slebbon

                                            I'm still in the process of just setting this up, but it appears that our clickable links for blades in the UCS Overview view in Solarwinds is showing the Blade's KVM IP address.  I've noticed that even though we have 2 chassis with 7 blades each (one is double-wide), however only blades 1 and 7 in chassis-1 show up as clickable (2 through 6 aren't).  I'm assuming this means that solarwinds UCS polling hasn't gotten IPs returned from UCS for the other blades?  Chassis-2 however has all 7 blades as clickable.  I wasn't involved in setting up our UCS and the documentation i have shows that we allocated 24 IPs into a "UCS Blades KVM Pool".  I'm not sure if that's a real thing yet or just something the guy setting it up called it.  If my information is correct though (it very well may not be) some of those blades in chassis-1 have been replaced with faster models since UCS was first installed years ago.  It seems very likely to me that some association or assignment in UCS Manager might not have been done at the time the blades were swapped, and this might be the reason that some of the blades aren't getting polled IP address info in Solarwinds?  But i know very little of this environment yet.

                                             

                                            What still confuses me is that this thread seems to indicate (and the fact that I'm seeing the KVM IP addresses) that the blades links in UCS Overview in solarwinds won't ever link to the Host OS (in our case ESXi), but other threads and the setup doc indicate that i should just add the blades clicking on the link and check 'poll for vmware' to get those blades added.  So i'm starting to wonder if there's some way in UCS to link the blade's IP that gets polled to say the OS host and we're just not setting UCS up correctly...causing solwarwinds to pick up the 'wrong' IPs?

                                              • Re: No, really, how DO you get UCS Blade Status?
                                                mesverrum

                                                No you are not doing it incorrectly.  Those blades can have vmware info polled on their regular NIC addresses, but it will not connect directly from this screen.  The UCS chassis view will list off the KVM addresses of the blades and if they got disconnected it would show them as being down, but the address associated with the OS (in this case ESX) is not visible from the UCS side and there is not a built in capability to link them within Solarwinds. You wouldn't try to poll for vmware on the KVM address because the ESX OS will not be answering those ports from that address, you have to point the vmware polls at the OS address.

                                                 

                                                As far as your blades that don't show up, I haven't worked on a UCS in a while and can't remember their management interface well enough to be sure but they may not be picking up KVM addresses automatically, id start by checking that they have those first.

                                                1 of 1 people found this helpful
                                            • Re: No, really, how DO you get UCS Blade Status?
                                              jest4kicks

                                              FWIW, I tried deleting and readding my UCS Manager IP as an ICMP node, polling for UCS.  Same thing I got before.  It shows the UCS FIs (as switches), with status and they're clickable, and it shows the chassis, also clickable, but none of the individual blades are clickable.

                                            • Re: No, really, how DO you get UCS Blade Status?
                                              lfransen

                                              Here is what mine looks like. After I reset the UCS Manager to the settings I mentioned earlier I was then able to click on each of the devices and add them separately. Capture.PNG

                                              • Re: No, really, how DO you get UCS Blade Status?
                                                lfransen

                                                the UcSManager is through the OS address and the blades come up as the blade management address when I click on them to add

                                                • Re: No, really, how DO you get UCS Blade Status?
                                                  dodo123

                                                  Right need some help on this

                                                   

                                                  We have two UCS one on each site with two FI each but only one seems to be showing up correctly in solarwinds see below.

                                                   

                                                  First off not sure why the one on the right doesnt bring back the response time, so went through the polling times and there were different so amended. This may take a while to update though.

                                                   

                                                  Compared all settings from the ucs manager and the fi's, now the only difference i can find that on the working one on the left: FI's they have ucs credentials that look like its being pulled from somewhere as the tick box is greyed out?

                                                  Working FI but cant untick??

                                                  Not working FI has no details or checked?

                                                   

                                                  So I think this is the issue the odd thing is the account thats used on the ucs is a different account to the FI so it isnt pulling it from there?

                                                   

                                                  Any help would be appretiated

                                                   

                                                  Thanks