I am trying to run tshark for an extended period of time to reassemble NFS requests that I capture using tcpdump (with pf_ring). I am streaming the pcap to standard out from tcpdump (using -w -) and into tshark (using -i -). It seems that tshark calls dumpcap to do the packet capture (even from stdin) and then dumpcap writes the packets to the /tmp directory and then tshark reads from there. The reading and writing to disk was a bottleneck for me so I specified a new location using the TMPDIR environment variable and set it to a tmpfs mount. This was successful in speeding up the tshark processing but once the tmpfs mount fills up, tshark crashes. First, I would like to confirm that this is the correct behavior. If so, I have two questions:
Thank you. asked 11 Dec '12, 13:17 notorious-pc... |
One Answer:
There is currently an old bug which says that tshark should not be using files when in this configuration (bug 2743). To answer your questions:
answered 12 Dec '12, 07:52 JeffMorriss ♦ |
Using the -b option when not writing out raw packets with the -w option does not work. The -b will only affect the explicit output. If there is no -w option, then tshark will just exit immediately.
Ah, right, makes sense. Oh well...