Wrong registered APs number from snmp query

[gdspa]

In last 2 weeks number of registered APs to our WLC9800, 17.9.4a, is wrong.

Now we obtain 305, but on wlc dashboard we have 273.

I don't find any bug, have anyone ever had this issue?

[Mark Elsen]

    - @gdspa                I would suspect a bug ; note that 17.9.x is getting older and EOL ; consider using
                                   a latest advisory release. You can also get a list of registered access points with :
                                               WLC  #   show platform  software dtls chassis active R0

  M.

[Jonatan Jonasson]

Is it possible that you're comparing the online/"operational" number from the WLC dashboard with the total number (online + offline + not-joined(disconnected)) from the SNMP query?

[gdspa]

I use oid 1.3.6.1.4.1.9.9.513.1.3.35.0

From wlc dashboard, joined APs are 273, not joined 79.

I know our version is old, we'll upgrade but we still have some 2700 APs, and I'd like to schedule upgrade after their renewal.

[Mark Elsen]

 - @gdspa                        17.12.6a   will work for       2700 models.

 M.

[gdspa]

Here https://www.cisco.com/c/en/us/td/docs/wireless/compatibility/matrix/compatibility-matrix.html it looks last version which supports 2700 is 17.12.3. Anyway, I was checking a version which fixes some bugs which cause random AP reboots.

[Mark Elsen]

  - @gdspa                   Good remark !  And it is often made; 17.12.6a will work with the 2700 models. the compatibility
                                    matrix was no longer updated after 17.12.3  because the later 17.12 versions where not
                                    sufficiently back tested on the older models , but that has no impact.

  M.

[gdspa]

Ok, thanks. I'll try to force a failover to check if snmp wrong value will be fixed.

[Mark Elsen]

  - @gdspa                  Ok, of course in HA SSO  different versions cannot coexist.    You can for instance download
                                   and deploy the 9800 VM (CL) for free; with up to 50 access points without complaints about licensing.
                                   That gives the opportunity to make a small lab  for testing with future versions; but if  done:
                                   Do not deploy it on the same subnet as the main controllers, because access points will always
                                   do a layer 2 capwap broadcast first; and you will loose  control on which controller they will use for
                                   joining

  M.

[gdspa]

I didn't share that snmp query wrong output started 1 day later after a reload of active wlc, this is why I'll try to change role, it should not change anything, but...

[gdspa · Accepted solution]

After forcing a failover, and some minutes of disaster (web interface of active wlc was showing first install wizard and all ssids were down even if by ssh I could see all the config), and a phisical reload of both WLCs, now snmp value is correct again. Cisco bugs are always in ambush...

[Mark Elsen]

    - @gdspa                  FYI     :    https://bst.cloudapps.cisco.com/bugsearch/bug/CSCwj73634
                                     Another good reason to abandon  the old  17.9.x  and use one of the 
                                     Known Fixed Releases mentioned in the bug report.
                                     In the meantime I would advise to copy the running configuration to
                                     an external repository before doing  an HA SSO  failover   (as an extra backup) ;
                                                              until you have upgraded
 

   M.