[Oisf-devel] Mem leaks

[Victor Julien]

On 10/18/2011 03:46 PM, Martin Holste wrote:
>> This will mean that any tcp session that doesn't have a packet within 10
>> seconds of the previous will time out. With interactive sessions like
>> irc/ssh/etc or slow responding servers this will mean you might loose
>> detection as tracking and thus reassembly will stop.
> 
> Yep, that's the trade-off we have to make for not dropping normal TCP
> traffic.  As I asked above, how can one tell when the emergency state
> is getting invoked, and how can one tell when the table is full
> despite the emergency state?
> 

The emergency state is currently printed to the screen or to your log if
you have set that up. I don't think there is another way of finding out
about it currently.

I think if the flow memcap is reached we're entering the emergency state
as well.

Definitely room for improvement here. How would you ideally be told
about this?

-- 
---------------------------------------------
Victor Julien
http://www.inliniac.net/
PGP: http://www.inliniac.net/victorjulien.asc
---------------------------------------------