[Snort-users] snort dropping 48%
josh.berry at ...10221...
Fri May 7 08:58:08 EDT 2004
Is it possible that the default rules with Snort aren't really catching
anything currently? And that your rulesets are which is why the
directories are created? Also, did you remove the -l switch from the
> I just performed a test. I'm running Snort 2.0.5 on RH Linux 8.
> If I startup snort with -N and the default rules enabled, then NO
> directories are created. If I then go into snort.conf and add my custom
> "my.rules", then run snort again with the SAME script, directories are
> then created!
> This looks like a bug in Snort as the rules shouldn't determine whether
> Snort creates the log directories. Thoughts from anyone?
> I'll play around with my custom rules to see if I can narrow the issue
> down further.
> -----Original Message-----
> From: Josh Berry [mailto:josh.berry at ...10221...]
> Sent: Friday, May 07, 2004 9:19 AM
> To: Sheahan, Paul
> Cc: snort-users at lists.sourceforge.net
> Subject: Re: [Snort-users] snort dropping 48%
> I have had no problems running snort on Gigabit once I tuned the system.
> One of the biggest differences was disabling the log directory creation
> with -N, and completely removing the -l switch. Using -l with -N
> make sense, isn't that like saying log, don't log?
> I also found information on tuning the Linux filesystem, and tuning the
> network cards. It helps to use a NIC capable of using NAPI. It sounds
> though you might have issues with the NIC drivers. If it is not the
> drivers you could try using the MMAP version of libpcap at:
> I also have a custom kernel that I stripped everything out of that
> completely necessary and then added performance enhancing patches for
> scheduler, the low-latency patches and pre-emptive patches.
> If you are running on a dual processor system you could set the CPU
> affinity for the Snort process.
>> Sheahan, Paul wrote:
>>>Also Snort STILL creates individual directories for
>>>each address it encounters. So many directories get created in reaches
>>>the Linux limit after a while and crashes Snort. I suppose Snort could
>>>be so busy with this that it may be contributing to the packet loss?
>> If you specify the -N switch it should not do any packet logging. I
>> tested this with `snort -d -l ./ -N -c /usr/local/etc/snort.conf'. It
>> generates the alert file , but not any packet logs, sounds like you
>> might not be using the -N switch properly (or the -N switch needs to
>> in a certain spot?). I could see how default packet logging could
>> kill a server that runs on gigabit though.
>> While this may contribute to it, it doesn't sound like the root of
>> problem though as you've previously tried logging binary format.
>> Sheahan, Paul wrote:
>>> The content rules are the issue, but it is still a mystery why old
>>> hardware and Snort version worked.
>> The real difference here is a amount of traffic snort needs to
>> Gigabit ethernet is a 10x faster than standard. Thats a lot of
>> What we really need is a response from someone who effectively runs
>> snort on a gigabit network. Can snort run "out of the box" on a
>> network efficiently (given decent hardware of course) or does it need
>> be tweaked to prevent major packet loss?
>> As for your current situation Paul, would it be feasible to share the
>> load between multiple sensors? Each sensor containing 100 of your
>> rules? That might work to get every packet on the wire without having
>> sacrifice some of snort's features for speed.
>> Just an idea. :)
>> This SF.Net email is sponsored by Sleepycat Software
>> Learn developer strategies Cisco, Motorola, Ericsson & Lucent use to
>> deliver higher performing products faster, at low TCO.
>> Snort-users mailing list
>> Snort-users at lists.sourceforge.net
>> Go to this URL to change user options or unsubscribe:
>> Snort-users list archive:
More information about the Snort-users