Memory leak in HiveOS 6.4r1?

  • 1
  • Question
  • Updated 3 years ago

I've been dealing with my Aerohive support company to help me resolving couple of problems I have at the moment (AP turning into Mesh Point without a reason and SNMP daemon on AP doesn't respond to queries).

To resolve the first problem I've been asked to roll back to HiveOS 6.2r1 and monitor one of my APs. I am bit reluctant to do so and I decided to ask if someone else has had this problem or seen it before?

So far I have three different APs with the 'Mesh Point' problem - 1x AP120 and 2x AP121. If I reboot them they are Ok (Portal mode) for few weeks (sometimes few months) and then they turn to Mesh Point again. Only reboot can change it back to Portal mode.

According to support person he can see numerous memory errors on the access point.

Is there a memory leak in HiveOS 6.4r1? Shall I roll back to 6.2r1?

Photo of Dariusz Chorzepa

Dariusz Chorzepa

  • 49 Posts
  • 1 Reply Like

Posted 3 years ago

  • 1
Photo of Mike Kouri

Mike Kouri, Official Rep

  • 1030 Posts
  • 271 Reply Likes
Dariusz,
I can acknowledge that in certain circumstances processes within HiveOS that allocate memory do not release it when they are done, leading to eventual memory exhaustion. These are not always trivial issues to find, much less fix, and this problem has plagued many releases of HiveOS. Sometimes, when we fix an issue in one release of code, the changes cause a different area of the software to begin exhibiting similar symptoms. And very often, these issues only expose themselves in certain client loads and traffic patterns.

Therefore, it is quite feasible that given your environment (client types and volume, traffic patterns and volume) you never exposed a flaw in 6.2r1 but did expose it in 6.4r1. It sounds to me like the support person's suggestion is legitimate.
Photo of Dariusz Chorzepa

Dariusz Chorzepa

  • 49 Posts
  • 1 Reply Like
Thank you Mike for your thorough and honest answer. I appreciate it. I have rolled back and I am monitoring the APs.
Photo of Bruce D.

Bruce D.

  • 2 Posts
  • 0 Reply Likes
Dariusz, 

Just found this thread researching the same issue: I have a couple AP121's on 6.4r1 that have been having random reboots and switching to Mesh Point mode, even though the eth0 link is up/up.  I'm wondering if it's the same issue... going to try a rollback as well, and report back in a couple days. 

-Bruce
Photo of Dariusz Chorzepa

Dariusz Chorzepa

  • 49 Posts
  • 1 Reply Like

Hi Bruce,

It sounds like the same problem I have. I rolled back to 6.2r1 and so far so good. The only reason why I wanted to use 6.4r1 is the ability to whitelist MAC addresses using CWP authentication but since I found out that  AP120 are not getting the same HiveOS version it is not so 'disastrous' for me to use older firmware  on AP121.

Please keep me posted if the problem is resolved for you with your rolled back version of HiveOS.

Thanks,

Dariusz

Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
As you still have AP120s in use, for consistency I would be minded to just run with an additional SSID rather than using MAC address whitelisting.
(Edited)
Photo of Nick Lowe

Nick Lowe, Official Rep

  • 2491 Posts
  • 451 Reply Likes
HiveOS 6.4r1d is now available if you wish to retest with this version. I am under the impression that there may have been a CWP related memory leak fix made in this build over 6.4r1a.
(Edited)
Photo of Dariusz Chorzepa

Dariusz Chorzepa

  • 49 Posts
  • 1 Reply Like
Thanks Nick. I will update the firmware on one of the test APs and see if the problem is resolved.
Photo of Bruce D.

Bruce D.

  • 2 Posts
  • 0 Reply Likes
Hey all,  quick update -- the two AP121's that were having issues on 6.4r1 have been 100% stable since downgrading to 6.2r1b.  Hope to get those memory issues fixed in a later release!

-Bruce
Photo of Dariusz Chorzepa

Dariusz Chorzepa

  • 49 Posts
  • 1 Reply Like

The APs (HiveOS v6.4r1d) were running Ok until yesterday and I lost ability to query 4 APs via SNMP and one turned into Mesh Point. They were running Ok for 21 days so it looks like the problem is still with the latest HiveOS :(