Hmmm.  Simple way, make your ntop host "see" many networks, then ping
every ip address on each network.  Or, remove the 1024? Host limit per
network, and create a /16 network and start pinging away.  Basically
just manifest a bunch of hosts - if you can manifest a bunch of sessions
as well, the would be better!

 

I'm still at 435MB with "only" 2,300 hosts and 7,500 sessions.  Roughly
same memory as when is was 30K hosts and 30K sessions.

 

Also, I THINK my cpu util issue MAY be due to a loop condition in
event.c - but not sure if it's your event.c or libevent event.c.  I'm on
1.4.11 libevent, but they have a 2.0 x release that's supposed to be
much better at various things.  If I restart this instance right now I'd
bet it would be < 100MB and < 5% cpu.

 

I'll send you some gdb output privately.

 

TIA!

 

Gary

 

 

________________________________

From: [email protected] [mailto:[email protected]] On Behalf Of
Luca Deri
Sent: Thursday, June 25, 2009 4:56 PM
To: [email protected]
Subject: Re: [Ntop] Race condition in 3.3.10?

 

Gary 

how can I reproduce this meMory issue?

 

Luca


Il giorno 25/giu/2009, alle ore 23.06, "Gary Gatten"
<[email protected]> ha scritto:

        Ps - I have trace level at 5 and nothing interesting in the
logs.
        
        Also, it *appears* as though memory isn't being freed. I have
a/v scanners that cause ntop to see upwards of 30,000 hosts for brief
periods every 4 hours. Obviously it must allocate memory for these, but
after they are purged the memory should be released right? I don't have
"purge cache" or whatever enabled - I can't get it to compile with that
option set.

        
________________________________


        From: [email protected] 
        To: [email protected] ; [email protected] 
        Sent: Thu Jun 25 15:54:37 2009
        Subject: [Ntop] Race condition in 3.3.10? 
        
        

        STILL trying to track down high CPU load since the upgrade to
3.3.x.  This morning one of my nTop instances that normally runs < 10%
(even on 3.3.10) jumped to about 50%.  It might go even higher but I
have another instance also running at 50% - it was 90% before this one
jumped.

         

        Anyone, the netflow thread is usually number 5 - and in this
case is as well.  The thread taking all the CPU is number 1, which is
typically the "main" process.  Thread 1 and 5 are usually the busiest,
but when thread 1 is taking roughly 10x the cpu as thread 5, something
is wrong.

         

        I'm thinking it was running along all fine and happy until some
set of conditions occurred that caused it to freak out - some sort of
race / locking issue?  Nothing changed "significantly" in my network to
account for the 5 - 10 fold jump in util for this instance.  Roughly the
same number of flows, pps, hosts, sessions, etc.  It was 5%, then bang -
50% and been there ever since...

         

        I CAN attach to this with gdb and look at stuff - if someone
smarter than me can assist.  I can make since of SOME of the gdb output,
but not having a strong development / debugging background is limiting
me.

         

        TIA!

         

        Gary

         

         

         

        "This email is intended to be reviewed by only the intended
recipient and may contain information that is privileged and/or
confidential. If you are not the intended recipient, you are hereby
notified that any review, use, dissemination, disclosure or copying of
this email and its attachments, if any, is strictly prohibited. If you
have received this email in error, please immediately notify the sender
by return email and delete this email from your system." 

        "This email is intended to be reviewed by only the intended
recipient and may contain information that is privileged and/or
confidential. If you are not the intended recipient, you are hereby
notified that any review, use, dissemination, disclosure or copying of
this email and its attachments, if any, is strictly prohibited. If you
have received this email in error, please immediately notify the sender
by return email and delete this email from your system." 

        _______________________________________________
        Ntop mailing list
        [email protected]
        http://listgateway.unipi.it/mailman/listinfo/ntop






<font size="1">
<div style='border:none;border-bottom:double windowtext 2.25pt;padding:0in 0in 
1.0pt 0in'>
</div>
"This email is intended to be reviewed by only the intended recipient
 and may contain information that is privileged and/or confidential.
 If you are not the intended recipient, you are hereby notified that
 any review, use, dissemination, disclosure or copying of this email
 and its attachments, if any, is strictly prohibited.  If you have
 received this email in error, please immediately notify the sender by
 return email and delete this email from your system."
</font>

_______________________________________________
Ntop mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop

Reply via email to