Spectrum WLC and Access Point Monitoring

[Brian Hunter]

​We have a Spectrum r10.3.2 DSS and fault-tolerant environment, and are attempting to monitor Cisco WLC managed access points.  We have about 60 WLC's and 2000 Cisco AP's which are modeled on a single landscape.  I've opened a case with support, but decided to post to the community in case some other users would have some useful input.
  One issue that I'm experiencing is that there are AP's which are known to be network unreachable, but are not alarmed in Spectrum until a manual "Ping" or "Poll" is executed, and just displaying a "Normal" condition in the meanwhile. So it seems that Spectrum is not ICMP polling these devices every 300 seconds as configured. (Poll Interval 300, Polling On, DCM Timeout 3000, DCM Retry Count 3).
  On the same Spectrum landscape there are pingable and snmp models which seem to have no monitoring issues. So it's frustrating when a user of Spectrum, like PC support, reports to us that Spectrum isn't reporting an issue when one exists.

[Brian Hunter]

​Perhaps I should have asked if other Spectrum users are having good results with WLC and Access Point monitoring.  My organization had tested the WLC Manager functionality when if first became available but could not use it due to various issues which supposedly have been overcome by r10.3.

[Silvio Okamoto]

Hi Brian,

I am not aware of an AP issue like this, even in 10.3.2 release.
I have seen duplicated APs due the same AP is assigned to multiples WLCs. APs were assigned a WLC for a test purpose and then they were renamed and assigned to another WLC.


------------------------------
Technical Support Engineer IV
Broadcom Inc
------------------------------

[Brian Hunter]

As I stated in a previous comment, it seems that the wireless access points are not being polled as expected.  I'll pick one for example.  An access point has a "Last Successful Poll" date of May 29, 2019, but has a condition of Normal.  I cannot however get an ICMP response from that AP when using a command line ping from my laptop computer.  So I do a manual poll from the Spectrum console and it's unsuccessful.  Subsequently, that manual poll triggered an event which created a critical alarm for that access point.  Why there was no critical alarm prior to the manual poll seems to be a big issue to me.  When I review the events, the event prior to this critical event of July 15 is a "Model XXXXX of type AccessPoint has been contacted" event with a time stamp of July 13.  I'll need some explanation of how I should interpret this behavior.  Why would a critical alarm for AccessPoint has stopped responding not be created until a manual poll of the AP from the Spectrum console is executed?

[Brian Hunter]

Yesterday, July 29, we deleted all of our 63 APs (which removed the 2044 APs). We did a re-discovery with AP modeling enabled. All APs and WLC's are modeled on one landscape, in the same container. (Spectrum 10.3.2) So this morning I executed a locator search for All APs. There were 209 APs showing normal condition and all had a Last Successfull Poll date/time of July 29, yesterday. I selected those and did a manual ping. Of the 209, there were 9 that failed ping. I repeated the ping about 10 minutes later, and the same 9 still had ping failures. The 9 APs now have DEVICE HAS STOPPED RESPONDING TO POLLS alarms as a result of these manual pings. They should have already had alarms sometime prior to this! It seems to me that the behavior is inconsistent. Over time, the number of APs having old Last Successful Poll date/times increases, and it appears that Spectrum is not monitoring them. If we cannot get the monitoring of APs to be stable and consistent, then it's no good for us. Resorting to manual creation of pingables for each of the 2044 APs is our only other option. I should add that the SSPerformance metrics look good, and so resource issues are apparent.

[Brian Hunter]

I've posted to this community thinking that someone may have ran into this issue, but perhaps not.  I've had a support case open for a while.  This is probably my last posting in regards to this problem, as it's not helpful.

Today, in morning checks, found that out of 2056 access points showing up in the All APs locator search, only about 45 have a "Last Successful Poll" date/time of July 31.  So, things seem to not be good here.

Somehow, even though I've been told that the access points are being modeled as pingables, they appear to be different than Spectrum Model Types of "Pingable".  The access points are of Model Type "AccessPoint", and are not being polled the same as pingable models.

Perhaps someone needs to review how the software is currently implemented to determine why the Spectrum Models of Type "AccessPoint" are not being polled and updated with proper status in Spectrum.