Re: Re: [opm] memory leak?

The file descriptors are alright. The system allows me
to use 65K FDs (this is the limit for the bot) and
this load is foreseen in the config file. The bot uses
about 600-800 FD in the evening and 60-120 in the
night. I use 2 scanners, one that scans 10
port/protocol pairs and another that scans 30-35 such
pairs. I use the extended scanner on limited networks
that has been proved to have proxies on strange port
numbers. The mashine the bot is running on can stand
this heavy load without much problems, despite the
female name ;-)

I don't think it is the scan queue that grows. If it
was the scan queue then during the night (when the
fdstats shows about 100 FDs), the bot must have
lowered its memory usage.. It was never the case. The
memory only increases and never decreases, forcing me
to restart the bot when it becomes about 500MB.

This situation is also persistent. It happends always,
regardless of the tweaks in the config file I have
done so far (so it's not the negcache as it was first
suggested).

Also.. in the night, when the load is not so big, the
bot almost does not increase in size. That is why on a
normal bot, that has about 3 to 10 connections per
minute, you're likely never to notice anything unusual
that might suggest a memory leak or some other
problem.

And yes, indeed, my tests with valgrind showed no leak
problems at all but still the memory valgrind was
using together with the bot was 400MB after about 17
hours of work.


Now, another 2 hours later from my previous letter,
the bot is using:
ongeboren@marla:~$ top | grep 7261
 7261 ongebore   9   0  508m 485m 9612 S  0.3 48.1  
9:14.69 bopm


Regards,
ongeboren


--- Andy Smith <[email protected]> wrote:

> On Mon, Jan 03, 2005 at 09:46:31PM +0100, Christian
> 'HERZ' Makowski wrote:
> > BOPM has imho no Memory Leak.
> > Look at your connection rate : (366.95/minute)
> 
> !!!!!
> 
> How many ports/protocols do you have in your conf?
> 
> If we assume you have 50 port/protocol combos and
> 1000 available file
> descriptors then you can scan 1000/50 = 20 clients
> at once.  Compare
> this with how many clients you are trying to scan,
> some 6 per
> *second*.
> 
> You can use the command "fdstat" to see how many
> file descriptors
> (of max) are in use.  If it's consistently at
> maximum then you have
> a problem.
> 
> As I already told you, scan queue will grow without
> limit and this
> was the first thing I told you to check.
> 
> -- 
> Andy Smith -- Occasional BOPM Developer And Support
> Monkey.  Please copy
> all BOPM support queries to the BOPM list, _not_
> just directly to me!
> If I've helped you with BOPM then please check my
> wishlist!
>
http://www.amazon.co.uk/exec/obidos/registry/23IJ4U7N4J3X9
> 

> ATTACHMENT part 2 application/pgp-signature 



__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com