Wireshark-dev: Re: [Wireshark-dev] Memory consumption in tshark
Date Index · Thread Index · Other Months · All Mailing Lists
Date Prev · Date Next · Thread Prev · Thread Next
From: Anders Broman <anders.broman@xxxxxxxxxxxx>
Date: Thu, 29 Aug 2013 14:58:32 +0000

 

 

From: wireshark-dev-bounces@xxxxxxxxxxxxx [mailto:wireshark-dev-bounces@xxxxxxxxxxxxx] On Behalf Of Evan Huus
Sent: den 29 augusti 2013 15:21
To: Developer support list for Wireshark
Subject: Re: [Wireshark-dev] Memory consumption in tshark

 

On Thu, Aug 29, 2013 at 5:32 AM, Dario Lombardo <dario.lombardo.ml@xxxxxxxxx> wrote:

On Wed, Aug 28, 2013 at 3:45 PM, Evan Huus <eapache@xxxxxxxxx> wrote:

Also, just as a useful reference point: if you run your normal tshark command (no valgrind) on the 1GB subset, what does memory usage of the process peak at?

 

It uses about 1,5GB of memory.

 

With the env vars you asked me to export, tshark can't run on the 1GB pcap file. I've reduced the dataset to 2 millions of pkts, and that's the output.

 

That's quite useful actually: it looks like the majority of the memory is being used to store address-resolution data from all of the DNS packets so that if those IP addresses show up later we can resolve them immediately (without having to ask the system name resolver).

It doesn't look like there's a way to disable this at the moment (I believe we still store the names even if name resolution is disabled), but it should be easy enough to fix. The add_ipv4_name and add_ipv6_name functions should probably be no-ops if all name resolution is disabled. Then simply passing the -n flag will greatly reduce your memory usage (though it won't yet).

If nobody has any objections or better ideas I will probably fix this on the weekend.

Cheers,
Evan

 

Currently also get_hostname() causes the IP to be added to the hash table presumably on the presumption it will be faster than doing ip_to_string().