Make keepalive interval configurable

[nhas]

Make keepalive interval configurable

[NHAS]

Oh yeah that is super annoying for you. I'll make this configurable with 0 being disabled.

[NHAS]

Sweet this is now on the unstable branch. The flag is --timeout, feel free to check it out (and close the issue if all is well)

[neben]

Wow, that was quick! Works great, but there are still the TCPKeepAlive messages being sent every 15 seconds, apparently from both sides.

10:46:28.646856 eth0  Out ifindex 386 02:42:0a:00:00:13 ethertype IPv4 (0x0800), length 72: 10.0.0.19.3232 > 10.0.0.10.46976: Flags [.], ack 73, win 22, options [nop,nop,TS val 2087557225 ecr 66711928], length 0
10:46:28.669283 eth0  In  ifindex 386 02:42:0a:00:00:0a ethertype IPv4 (0x0800), length 72: 10.0.0.10.46976 > 10.0.0.19.3232: Flags [.], ack 105, win 3790, options [nop,nop,TS val 66713450 ecr 2087542031], length 0
10:46:28.882232 eth0  In  ifindex 386 02:42:0a:00:00:0a ethertype IPv4 (0x0800), length 72: 10.0.0.10.46976 > 10.0.0.19.3232: Flags [.], ack 105, win 3790, options [nop,nop,TS val 66713472 ecr 2087542031], length 0
10:46:28.882256 eth0  Out ifindex 386 02:42:0a:00:00:13 ethertype IPv4 (0x0800), length 72: 10.0.0.19.3232 > 10.0.0.10.46976: Flags [.], ack 73, win 22, options [nop,nop,TS val 2087557461 ecr 66713450], length 0

Is there a way to disable this like with sshd_config's TCPKeepAlive option?

[NHAS]

Oh wild, I didnt even think of that, apparently this is a feature added in go 1.13, I think I can disable this server side. But Im not keen on modifying the client.

[NHAS]

Sweet, --timeout should now also allow you to disable/configure that. I personally havent tested that it works, so confirmation would be good

[neben]

It somehow works, but not well. The client sends two keep-alive packets at the same time, every 15 seconds, which are confirmed by the server. This looks like a bug to me.

11:00:34.866020 IP 127.0.0.1.64391 > 127.0.0.1.3232: Flags [.], ack 1134, win 6361, length 0
11:00:34.866044 IP 127.0.0.1.64391 > 127.0.0.1.3232: Flags [.], ack 1134, win 6361, length 0
11:00:34.866055 IP 127.0.0.1.3232 > 127.0.0.1.64391: Flags [.], ack 1664, win 6353, options [nop,nop,TS val 780324830 ecr 3834378403], length 0
11:00:34.866060 IP 127.0.0.1.3232 > 127.0.0.1.64391: Flags [.], ack 1664, win 6353, options [nop,nop,TS val 780324830 ecr 3834378403], length 0
11:00:49.927164 IP 127.0.0.1.64391 > 127.0.0.1.3232: Flags [.], ack 1134, win 6361, length 0
11:00:49.927192 IP 127.0.0.1.64391 > 127.0.0.1.3232: Flags [.], ack 1134, win 6361, length 0
11:00:49.927205 IP 127.0.0.1.3232 > 127.0.0.1.64391: Flags [.], ack 1664, win 6353, options [nop,nop,TS val 780339891 ecr 3834378403], length 0
11:00:49.927208 IP 127.0.0.1.3232 > 127.0.0.1.64391: Flags [.], ack 1664, win 6353, options [nop,nop,TS val 780339891 ecr 3834378403], length 0

[NHAS]

Nah thats intended. There isnt a way of configuring this for clients, so the client still does the keep alive back to the server.

[neben]

Yeah, but the question is why does it do it twice?

[neben]

Nevermind, it works fine (was an issue with tcpdump counting the packets twice)! I can live with the client sending a TCPKeepAlive every 15 seconds. Thanks again for the great work!

[NHAS]

Yeah, but the question is why does it do it twice?

Huh! Honestly I didnt even pick that up from what you sent, must be a bug in the unix kernel mac uses (or potentially a feature :P ).

In hindsight I'll make the default keepalive on the client an hour or so. Will help you reduce cost even more!

I would be keen to hear a little bit more about how you're using it, and if there is anything else you'd want.

[NHAS]

Sweet, clients now have a 2 hour keep alive