12.2(58)SE2 on a 2960 cause memory leaks

[pasztor.richard]

Hi all,

I had a very disappointed experience today  during the planned upgrade on my 2960 to IOS version 12.2(58)SE2 after business hours. We have a small office with 3x 2960 and a couple of 2950. There is a single 2960 called as the "central" as its the STP root, and most other switches are directly connected to the central. The remaining switches are conected to one of the switches conneting to the central.

I tested the aforementioned IOS on 2x 2960 las t week, but did not upgrade the central one. The 2x switches ran fine for a week, so I decided to upgrade the central one: basically it was a disaster. After the restart, the central became immediately unavailable via SSH (though admin IP pingable). I connected to the console, and realized that even console does not work, as I received the error message: "Low memory", and the console session could not be started. I restarted the switch once again (had to cut the power) and this time I managed to connect via the serial: the free memory was approx 2 MB, though its was over 5 MB if running 12.2(55)SE1. After 1-2 mins, free memory decreased down to 1.8 megs. SSH was still not functioning. At this time, fearing of a sudden memory leak during the night, I had to rollback to the previous version.

My config is quite simple: RSTP root, DHCP snooping, 1 LACP etherchannel, only 5x VLANs with VTP server, MOTD and BANNER (yeah I know its a risky feature since the previous IOS) I wonder what can be the problem here?

I have attached the running config of the central 2960 (removed some sensitive entries)

[Jonathan Hall]

Hi Leo,

The 2960-S has double the memory of the 2960 so I can see why you're not seeing this issue.

It looks like 12.2(58) and 15.0(2) are just too memory hungry for a 2960. I'll stick with 12.2(55) until I can replace my 2960 switches.

Thanks,

Jonathan

[Leo Laohoo]

Processor Pool Total:   21751188 Used:   18857412 Free:    2893776
I/O Pool Total:    4194304       Used:    2380052 Free:    1814252
Driver te Pool Total:    1048576 Used:         40 Free:    1048536

Oh shoot!

Plain 2960 ... Ok, here's mine.

Forgot to ask you, are you seeing this issue on a 48-port PoE switch?

[Jonathan Hall]

Leo Laohoo wrote:
 
Forgot to ask you, are you seeing this issue on a 48-port PoE switch?

WS-C2960-24TT-L is the only model of 2960 I currently use, so no 48-port PoE to check I'm afraid.

[ashishpoyekar]

Dear All,

I search for cisco 2960 S series stable IOS & login to cisco site & found below that cisco suggested me 12.2.55 SE9 . More details check below attached file.

Thanks

[Chris Fink]

I had this problem too. Turning off macro auto monitor seemed to help (no macro auto monitor). Nevertheless, I have downgraded all of my 2960 plain switches to 12.2(55)SE8.

[Ankur Atri]

WS-C2960-48PST-L memory leak issue with Version 15.0(2)SE4. Will there be a relief for us. On some IOS Hulc LED won't let us live and now this mermory leank. Which IOS is stable. We have plenty switches.

And this one also has the HULC LED issue.  

Cisco IOS Software, C2960 Software (C2960-LANBASEK9-M), Version 15.0(2)SE4, RELEASE SOFTWARE (fc1)
INND-S-A-POE-17#sh proc cpu sor | ex 0.00
CPU utilization for five seconds: 61%/31%; one minute: 62%; five minutes: 67%
 PID Runtime(ms)     Invoked      uSecs   5Sec   1Min   5Min TTY Process
 133     2528959      217020      11653 15.33% 15.39% 16.61%   0 Hulc LED Process
  87      376601      216612       1738  4.79%  5.18%  4.69%   0 HLFM address lea
  74      118027      372358        316  1.11%  1.48%  1.26%   0 Fifo Error Detec
 201       46260      113931        406  0.95%  0.63%  0.53%   0 Spanning Tree
 230       42708       19032       2244  0.79%  0.43%  0.40%   0 Marvell wk-a Pow
 102       19929       73009        272  0.63%  0.30%  0.21%   0 hpm main process
  70       97236      265213        366  0.47%  0.60%  0.56%   0 RedEarth Rx Mana
  89       15734      218230         72  0.31%  0.14%  0.10%   0 HLFM address ret
   2        1055        1758        600  0.15%  0.05%  0.01%   0 Load Meter
 142       17479        1769       9880  0.15%  0.16%  0.15%   0 HQM Stack Proces
  69        9643       67700        142  0.15%  0.07%  0.06%   0 RedEarth Tx Mana
 123        7116       42943        165  0.15%  0.05%  0.05%   0 Hulc Storm Contr
 165       30861       87697        351  0.15%  0.28%  0.19%   0 UDLD
 143        9724        3758       2587  0.15%  0.10%  0.07%   0 HRPC qos request
  49        4851        8821        549  0.15%  0.07%  0.03%   0 Per-Second Jobs

[jonhill]

We've just hit the memory leak issue on 12.2(58)SE2 and done some testing with 15.0(2) and found the bug again but also with other users recommendation downgrading to 12.2(55)SE9.

What I find interesting is that Cisco haven't updated 12.2(58)SE2 since 2011 but there's been multiple updates to 12.2(55) with the latest being this month and takes it to SE9, to me this speaks volumes on which firmware train to stick with.

[thiago.doni]

Hi Everybody

I read all the posts and still do not know what's the most stable version. 

I saw that the latest is 15.0.2-SE6 (ED). 

Has anyone tested this version?? 

What is the recommendation? 

Switch WS-C2960-S-48PST

Tks

[jonhill]

I've downgraded our 2960's to 12.2(55)SE9 at a recommendation from someone at Cisco.

[Parvesh Paliwal]

The only thing that I am curious to know is - It is working fine for other switches....for more than a week

[jonhill]

We were using 12.2(58)SE2 for the best part of 2 years before we started getting any issues then all of a sudden not just a couple of switches we had probably around 20 - 30 switches all start playing up over the course of a month. 12.2(55)SE9 so far is stable and haven't had any issues.