GUNNER can't (see and/or gun down?) jobs abandoned with TCPI status

[dseagrav]

If a job is started from a TCP connection which is then abandoned instead of closed (typical of port scanners, but maybe also network lossage), it will get left in the TCPI state:

 10 ___010 TELSER SYN027     TCPI    ?         4   3   0%

GUNNER either cannot see these jobs or cannot gun them down, and if enough of them accumulate the machine will become inaccessible due to exhaustion of network connections.

Edit: This behavior has only been confirmed on physical hardware, I would have to do a non-trivial amount of network reconfiguration to expose an emulator to the same network conditions as the physical machine.

[eswenson1]

Thanks for entering this report. This drives me nuts on my ITS systems. I get lots of inbound connections from the internet (hackers) and consequently a lot of these jobs hang around. I kill them every day with PEEK, but it would be good to update GUNNER to kill those that hang around without associated HACTRNs.

[larsbrinkhoff]

Does this apply to the Dynamic Modeling GUNNER, or the MC Trivial GUNNER? In any case, I suggest it would be good if both can handle this.

Also, how about abandoned Chaosnet and ARPANET jobs?

[dseagrav]

I have only tried this with the DM gunner. I don't have a means of testing NCP ARPANET with my present configuration. I will test Chaos this weekend and let you know what the results are.