WatchdogSec in systemd script kills tinc

[jvasile]

I've been running tinc since the mid aughts and it's been amazingly easy and stable. Thank you for such useful software!

This bug report is about the systemd scripts. The WatchdogSec=10 line causes systemd to kill tinc when the watchdog fails. Then systemd restarts the failed tinc. Then it kills it. This cycle continues indefinitely and makes tinc unusable.

Some effects of the watchdog killing tinc frequently:

• Every few seconds, the connection to another tinc node on the LAN dropped for about 7 seconds before resuming. This doesn't happen when connecting without tinc.
• Repeated runs of systemd status tinc@foo shows active, then not, then active, then not as tinc restarts.
• Pinging a tinc node results in packet loss of 30% to 75%. This doesn't happen when pinging without tinc.
• Running tincd from the commandline works fine but running via systemd produces these symptoms.
• Removing the watchdog line causes tinc to behave as usual when run via systemd.

I haven't worked with systemd watchdog timers, but my understanding is that tinc should be satisfying the timer somehow. Clearly it isn't. I'm not sure if the line should be removed or the timer satisfied somehow.

This issue arose on 1.1pre18 and also 1.0.36. Everything is run on a current Debian testing box. My network topology is just a big soup of peer nodes all talking to each other. My tinc.conf has Name and ConnectTo lines and nothing else.