[systemd-devel] systemd "start request repeated too quickly, refusing to start" kills "Restart=" functionality

[Lennart Poettering]

On Sat, 28.05.11 09:46, Mike Kazantsev (mk.fraggod at gmail.com) wrote:

> Good day,
> 
> I've been running an ssh tunnel with keepalives as systemd service with
> "Restart=always" option (since network connection there goes down quite
> frequently), and it worked fairly well until this happened:
> 
> May 25 20:41:47 daemon.info<30> systemd[1]:  net-rathole.service holdoff time over, scheduling restart.
> May 25 20:41:47 daemon.warning<28> systemd[1]:  net-rathole.service start request repeated too quickly, refusing to start.
> 
> After the last message systemd just stopped restarting the service,
> which I think isn't supposed to happen.
> 
> Is this really a bug or Restart= should work this way?
> 
> Machine is running systemd 26 or 27, unfortunately I can't access
> machine via ssh (because of the service in question, got syslog lines
> mailed in the report) to get exact version.
> RestartSec= option isn't set there, so I guess default 100ms holdoff is used.

Hmm, so currently if a restart fails due to ratelimiting we never try
again regardless whether Restart=always is set. 

You have a point in wondering about this behaviour and we probably
should fix this to restart the service after the ratelimiting period is
over.

I have added this to the todo list now.

Lennart

-- 
Lennart Poettering - Red Hat, Inc.