You need more memory. We do not recommend routinely using swap space at all - when
ntop walks it's structures to compute throughput et al, it will thrash.
Host Hash counts
Actual Hash Size.....69632
Stored hosts.....263069 [377 %]
^^^ > 100% is a dead giveaway
80% of a /20 is around 3200 LOCAL hosts, plus the 2, 3, 30, 500 that each of them is
exchanging information with (I've been told that a busy Kazza user can have sessions
w/ 5000! remote hosts). Your hash size is 69K.
It looks like ntop isn't properly handling a failing malloc() - I'll have to check
around for that, so it fails gracefully. Still, the answer is more memory.
It does take a while to free memory - it has to walk all the structures and free each
of the individually allocated parts. Plus, depending on the OS and library version,
it may be consolidating freed areas for reuse or to return to the OS in large chunks
(mmap)...
-----Burton
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of
listuser
Sent: Thursday, January 30, 2003 2:22 AM
To: [EMAIL PROTECTED]
Subject: [Ntop] Ntop crashing and memory usage
Hello all,
I am using ntop to observe a /20 network which is nearly full, ie approx 80% hosts are
up. (it's a cable ISP). my ntop box is a P4 1.7GHz 512 MB ram, 512 swap. My ntop is
compiled from the jan 26 (or 27) snapshot.
When ntop runs it takes up memory progressievly untill the swap becomes 0 (ram has
about 1M free), and then the process is killed. In between I get messages like
ntop[12890]: WARNING: releaseMutex() call with an UN-LOCKED mutex [address.c:347]
VPN ntop[12889]: WARNING: releaseMutex() call with an UN-LOCKED mutex [hash.c:150]
Another interesting observation was that when I stopped ntop running with 0 swap from
web interface it took about half hour to free the memory
_______________________________________________
Ntop mailing list
[EMAIL PROTECTED]
http://listgateway.unipi.it/mailman/listinfo/ntop
