%SYS-2-MALLOCFAIL on 2960x

[Joris Deprouw]

Hi all,

We have a 6880 vss setup with more than 20 2960x access switches connected to it with a port-channel.

Some time ago our network had major issues. Most of our access switches were no longer reachable, the access switches on which we could log on we saw very high CPU load. When we shut down 1 link of the port-channel to the vss, the cpu went down again, and normal operations resumed.

The IOS on the access switches is 12.2(2)E5. Is it the DHCP snooping process which was causing the memory issue?

In the log of the switches we saw the following messages:

%SYS-2-MALLOCFAIL: Memory allocation of 1692 bytes failed from 0x24C79E8, alignment 0

Pool: I/O  Free: 972  Cause: Not enough free memory

Alternate Pool: None  Free: 0  Cause: No Alternate pool

-Process= "HULC DHCP Snooping Process", ipl= 0, pid= 293

.....

Thanks,

Joris

[Julio E. Moisa]

Hi Joris,

I recommend this link:

https://supportforums.cisco.com/document/22301/user-receives-sys-2-mallocfail-messages-due-memory-leak-or-fragmentation-issue

Hope it is useful

:-)

[Joris Deprouw]

Thanks Julio,

br,

Joris

[Richard Burts]

Clearly this is an issue with memory on the access switches. The link suggested by Julio is quite old (especially as evidenced by its repeated references to Blaster and Nachi worms) but it has good suggestions. If the original poster can confirm that these access switches have been running the same version of code for a while, then I believe that we can discount the possibility that the switches do not have the required minimum amount of memory to run that version of code.

I was initially drawn to the possibility of a memory leak being responsible for the problem. But the statement that when one link in the port channel was shut down then normal operations resumed makes me believe that it was not a memory leak.

So the other suggested cause of the issue is memory fragmentation. We do not have enough information about the environment to be sure but I am guessing that something about that link in the port channel was causing some kind of processing that created fragmentation of memory.

HTH

Rick

[Leo Laohoo]

The IOS on the access switches is 12.2(2)E5.

Uhhhh ... that is not a valid IOS version number.  Kindly post the complete output to the command "sh version" as well as "sh logs".  I/We want to see the complete error message(s).