[c-nsp] Memory leak on in latest 12.0S on GSR?

[Sascha Pollok]

Me again,

>> These occur due to memory leaks in the BGP process.
>> Disabled BGP sessions can cause memory leakage and the type of CPU issues 
>> you are seeing.
>
> there are currently none on this box.
>
>> Solution is upgrade to latest 12.0(33)S train.
>
> The PRP-1 is on 12.0(33)S7. Quite new I guess.
>
>> Work around is periodic reloads and eliminating down/shutdown bgp sessions.
>> I am sure there are other bugs but not sure what all of them are.
>
> So I guess I am more likely hitting one of those other bugs :-(

I did some digging and found, that "show mem dead" is awfully long. Feels 
like millions of lines (could be ten thousands of lines, too). Most of 
them refer to a process that I can not find anything about:

07406718 0000000040 074066C8 0740676C 001  -------- -------- 001FF434 acct periodic timer

"acct periodic timer" - doh! Anyone with an idea what that is?

It is only 40 Bytes or so but it sums up to 90 MB or so as dead memory.
Could it be that the "check heaps" process is somehow related to dead 
memory and that those dead fragments are the reason for the CPU hogs
that we are seeing during Check Heaps?

Any idea is much appreciated.

Thanks
Sascha