SNMP information for CPU / Memory

  • 7
  • Question
  • Updated 3 years ago
  • Answered
My network engineer is trying to poll cpu and memory stats on AP330 access points via snmp, but is unable to find the OID for those. Are there SNMP statistics for those 2?
Photo of Van Jones

Van Jones

  • 75 Posts
  • 4 Reply Likes

Posted 5 years ago

  • 7
Photo of Mike Kouri

Mike Kouri, Official Rep

  • 1030 Posts
  • 271 Reply Likes
Van,
Sorry, but I don't believe so. You can see the full set of MIBs we understand by navigating within HiveManager to Home > Administration > Auxiliary files > MIB files.
Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
The MIB files are available from within HiveManager. I would not advise using a third party source as they will not version.
Photo of Fabrício Soares Jesus

Fabrício Soares Jesus

  • 6 Posts
  • 0 Reply Likes
Ok Nick, tell me exactly where I find the corresponding OID processing, memory and connected clients?
Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
Processor and memory usage is not supported at this time.

If you look at the MIB files in a text editor, you can see what can be currently queried about connected clients.

Please do not spam with the same messages posted repeatedly.
Also, if you are not following on from a message with a related comment, please do not place an unrelated comment beneath it. Rather, add a new reply.
(Edited)
Photo of Fabrício Soares Jesus

Fabrício Soares Jesus

  • 6 Posts
  • 0 Reply Likes
obrigado Nick, e clientes conectados?
Photo of vince

vince

  • 1 Post
  • 1 Reply Like
Hi Nick,

Please do you have any news about the snmp trap for cpu and memory?
Thank you
BR
Photo of Van Jones

Van Jones

  • 75 Posts
  • 4 Reply Likes
Please tell me that this can be added. We are experiencing issues with our access points and a key indicator is high cpu. We need to be able to monitor cpu and memory usage over time with Cacti so that we know there are problems before our customers complain.
Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
There's probably a way to get this information via SSH, documented or undocumented. I'll have a poke tomorrow.
Photo of Van Jones

Van Jones

  • 75 Posts
  • 4 Reply Likes
We were able to pull the information into Cacti via telnet, but I have feature request FR-1358 created to 'create and publish OID's for AP CPU and Memory levels'.
Photo of Jorge Carrizosa

Jorge Carrizosa

  • 1 Post
  • 0 Reply Likes
This definitely would be useful.
Photo of Ed

Ed

  • 2 Posts
  • 0 Reply Likes
More robust SNMP OID choices for gathering data is needed
Photo of Joe

Joe, Product Management

  • 13 Posts
  • 2 Reply Likes
I am putting together a requirement for this.  While I am at the MIBs, what other useful parameters we should look into, especially at the system level?
Photo of Fabrício Soares Jesus

Fabrício Soares Jesus

  • 6 Posts
  • 0 Reply Likes
Hello Joe,
You have the OID that monitors the amount of clients? informs number of clients connected, you know?
Photo of Tony

Tony

  • 5 Posts
  • 0 Reply Likes
Yes - The existing OID is actually a collection of 'STRING' values which you then have to count to produce a numeric value.

I was just asking to add this 'Numeric Count' to make it easier to monitor for people wanting to monitor and alert on the number of connections per AP.
Photo of Johan van der Ven

Johan van der Ven

  • 1 Post
  • 0 Reply Likes
Hello Joe, are there any Cacti templates available? I would very much like to know. Reporting of Aerohive is great, but sometimes long-term graphs are greatly appreciated
Photo of Eric Duesterhaus

Eric Duesterhaus

  • 15 Posts
  • 5 Reply Likes
This doesn't related to APs, but for switches, they should really have individual interface monitors as well as CPU/Mem OIDs.
Photo of Fabrício Soares Jesus

Fabrício Soares Jesus

  • 6 Posts
  • 0 Reply Likes
Searching and querying myhive-auth.aerohive.com, found the item (Home> Administration> Auxiliary Files> Files MIB) that helps a bit, with great effort managed to identify some OID's, for example: 
1.3.6.1.4.1.26928.1.1.1.2.1.2.1.3 - Tells the hosts that are connected. 
1.3.6.1.4.1.26928.1.1.1.2.1.2.1.4 - Informs ips by dhcp distributed to customers 
1.3.6.1.4.1.26928.1.1.1.2.1.2.1.1 - Determine the MAC Address of the connected clients
Photo of Tony

Tony

  • 5 Posts
  • 0 Reply Likes
I would like to suggest a simple 'Clients Connected' count, TimeZone settings, current AP Time.
Photo of Joe

Joe, Product Management

  • 13 Posts
  • 2 Reply Likes
Answering a few questions here:
* Fabricio, I plan to add connected client count, but may not be in the same mib.  I will see where it is most appropriate.
* Johan, I don't believe there's Cacti templates available yet.
* Eric, for interface, we have the IF-MIB, not sufficient?
* Ratfink, current AP time refers to uptime or just the current time relative to timezone?  Curious what you want that for?  Worry time is not synced?
Photo of Tony

Tony

  • 5 Posts
  • 0 Reply Likes
Actually - The 'Local' AP Time was an idea that popped up during our ACWP course yesterday.  It was suggested that it could be a way to make sure that the TimeZone settimgs were matching physical location of the AP
Photo of Joe

Joe, Product Management

  • 13 Posts
  • 2 Reply Likes
Thanks.
Photo of Rob

Rob

  • 42 Posts
  • 5 Reply Likes
This is a great idea as we had some issues with bad NTP coming from our router but only affecting some APs... We were getting complaints that people would move to certain areas in the building after being authenticated and the network would kick them off. I just happened to check the time on one of the bad APs and noticed it was off by a few years! Needless to say we didn't know the root cause of our bad time at so the local tech had to check all 300 APs and note which ones were having issues... He was not happy.
Photo of Tony

Tony

  • 5 Posts
  • 0 Reply Likes
Rob, This is exactly the reason we were discussing this as a monitoring option :)
Photo of Fabrício Soares Jesus

Fabrício Soares Jesus

  • 6 Posts
  • 0 Reply Likes
Joe, 

Conseguiu a OID para processamento, memory e clients conectados?

Photo of Eric Duesterhaus

Eric Duesterhaus

  • 15 Posts
  • 5 Reply Likes
Thanks for the answer  - I've imported all of the MIBs from hivemanager into our monitoring server (PRTG), and the interface monitors were working with our APs, but for some reason not with our switches.  I'll give it another shot today.
Photo of Joe

Joe, Product Management

  • 13 Posts
  • 2 Reply Likes
Thanks, Eric.  It should work.  If not, it should be a bug.
Photo of Fabrício Soares Jesus

Fabrício Soares Jesus

  • 6 Posts
  • 0 Reply Likes
Eric, 
Good afternoon, you know which specific OID processing, memory and clients connected?
Photo of Joe

Joe, Product Management

  • 13 Posts
  • 2 Reply Likes
Fabricio,

I think you might have the answers already.  None of the three you listed is supported in MIBs yet.  I am putting together the requirement right now but don't have a timeframe at the moment.

Joe.
Photo of JimmyBoJingle

JimmyBoJingle

  • 9 Posts
  • 3 Reply Likes
We are having issues with WIPs in our network, and an AH agent suggests it could be due to CPU spikes. It's a little frustrating the AH's don't support something as common as CPU polling via SNMP, because then I could tell in moments if that is really part of the problem or not.

Please add this!
Photo of Eric Duesterhaus

Eric Duesterhaus

  • 15 Posts
  • 5 Reply Likes
At risk of going off-topic - do you have dynamic airtime scheduling enabled?  We did, and it was causing huge and frequent CPU spikes in our environment (bad enough that they were symptomatic and causing problems with more than just WIPS).

We haven't bothered to pursue a case to actually fix it yet, because we didn't have a need for the feature.

Back on topic - CPU polling on APs would be great!
Photo of Stephan Hughson

Stephan Hughson

  • 8 Posts
  • 2 Reply Likes
Sorry to bump this up :-). We really need Load Average, CPU and memory monitoring via SNMP of each AP.
Has any work been done on this behind the scenes? It would be really helpful.

Or if the API could be extended then that would be even better. Thanks.
(Edited)
Photo of Brett Monroe

Brett Monroe

  • 2 Posts
  • 0 Reply Likes
I hate to poke this topic when so many have before me, but we could use CPU and memory monitoring via SNMP in our environment too.  

I think the big frustration is that we see this conversation being replied to in the past 9-12 months, but no further updates.  It would be nice to see some status feedback from Aerohive to the various requests above.  

I feel that posting this comment is equivalent to talking to the wall.  :)
B.
Photo of Mike Kouri

Mike Kouri, Official Rep

  • 1030 Posts
  • 271 Reply Likes
Stephan, Brett, (and other lurkers on this thread),
I know it may feel sometimes like you're talking in a vacuum, but I am pleased to announce that we WILL have support in the HiveOS 6.6r1 release now wrapping up development and releasing later this summer for the following OIDs:
  • CPU utilization (1.3.6.1.4.1.26928.1.2.3)
  • Memory utilization (1.3.6.1.4.1.26928.1.2.4)
  • Client Count (1.3.6.1.4.1.26928.1.2.9)
Photo of Brett Monroe

Brett Monroe

  • 2 Posts
  • 0 Reply Likes
Great to hear Mike, thanks for the feedback!
B.
Photo of Dianne Dunlap

Dianne Dunlap

  • 75 Posts
  • 15 Reply Likes
I have posted a Cacti template to track Aerohive clients, cpu, & memory in the Cacti forum:
http://forums.cacti.net/search.php?keywords=Aerohive&terms=all&author=&sv=0&sc=1&...
Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
At a first glance, the memory graph looks to be inaccurate in:

http://forums.cacti.net/viewtopic.php?f=12&t=54736&p=257232&hilit=Aerohive#p257232
(Edited)
Photo of Dianne Dunlap

Dianne Dunlap

  • 75 Posts
  • 15 Reply Likes
The memory is based on a very un-busy access-point and reflects "show memory" which seems to remain pretty constant.  If there's a different 'show'  for memory which should be graphed, we can do that!
Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
Look at the units on the left hand side of the memory graph. The access point that you are sampling has megabytes of RAM, yet the graph shows kilobytes as being used. I think you've got an issue with your unit mapping/conversion therefore.

In answer to your question, there is a command that gives more detail:
console page 0
show memory detail
console page 22
That would therefore give you more things to show on a graph:
Aerohive
Total Memory:        252348 KB
Kernel Free Memory:  115192 KB
Used Memory:         105188 KB
Buffers Memory:        6728 KB
Cached Memory:        43960 KB
Tmpfs Memeory:        11992 KB
Total Free Memory:  147160 KB
(Being pedantic, the output is wrong from HiveOS there as there's no such thing as kelvin bytes (KB), it should be a kB... or even more accurately KiB, as it's in binary - a power of 2.)
(Edited)
Photo of Dianne Dunlap

Dianne Dunlap

  • 75 Posts
  • 15 Reply Likes
Good catch on the graph scale - fixed that and will repost within the next day:


Do any of the other 4 memory variables seem useful from a support standpoint?
Photo of JimmyBoJingle

JimmyBoJingle

  • 9 Posts
  • 3 Reply Likes
It wouldn't hurt to have all those memory values accessible via SNMP OIDs, then admins can choose if they want to poll against all of them, or just the basic 3 you have shown. Having those extra data points could help diagnose intermittent or difficult issues - or at least prove conclusively its not related to memory usage in any way, which AeroHive tech support sometimes suggests.
Photo of Ed

Ed

  • 2 Posts
  • 0 Reply Likes
YES! Pease give us more SNMP data. Admins rely on more than just one tool. Hivemanager is good but it can't (and shouldn't) be everything. Spend some time on SNMP enhancement.

If you access it for data in Hivemanager, we want to access it in SNMP too.
Photo of Dianne Dunlap

Dianne Dunlap

  • 75 Posts
  • 15 Reply Likes
Fixed horizontal scaling and added all memory values to Cacti template at http://forums.cacti.net/viewtopic.php?f=12&t=54736&p=257343#p257343