initctl -j status had the state of a service but little of how it is
set up. A program reading it could not tell how the service is
stopped, what scripts it runs, or how it signals readiness.
Add the settings: stop, reload, notify, remain, scripts, directories,
cgroup-config, rlimits, log or tty, provides, conflicts, if, groups,
capabilities, pam, name, id, protected, and nowarn. The restart
object also gets sec, respawn, and oncrash, and the state Finit keeps
for its own use goes in internal:
"stop": {
"signal": "SIGTERM",
"timeout": 3,
"scope": "pgroup"
},
"scripts": {
"exec-stop-post": { "command": "/bin/true", "timeout": 7 }
},
initctl reads the status over D-Bus first, so each setting is also a
property of org.finit.Service1. The new test checks that the JSON is
the same over D-Bus as over the initctl socket.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
initctl -j status is for other programs and should have all there is,
the plain view is the one kept short. It was the other way around:
the restart counter and restart-max were only in the plain view, and
the restart delay of restart-max-sec was not in the JSON either.
Add a restart object:
"restarts": 7,
"restart": {
"count": 4,
"max": -1,
"delay": 160,
"limit": 300
},
Also add manual and started, and give starts for all services, not
only those with manual-start. The restarts total stays where it is.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With restart-max-sec the delay before the next restart varies, and
nothing showed what it was or how far it could grow.
Show both in initctl status, for a service that has the setting:
Restarts : 7 (4/-1)
Restart in : 160 sec (max 300)
Add the D-Bus properties RestartDelay and RestartDelayMax for it, so
the status is the same when initctl goes over D-Bus.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
restart_sec: was the only setting in snake_case. Every other option
of the line-based format is in kebab-case, and so is the block format,
restart-sec = SEC.
Take restart-sec: and keep restart_sec: as an alias, so a .conf from
an older Finit loads as before:
service restart-sec:10 foo -- Foo daemon
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The delay between restarts of a crashing service is fixed: two
seconds, then five, or restart-sec if that is longer. A service that
should be retried quickly at first and then less and less often, e.g.
a modem dialer, cannot be set up.
Add restart-max-sec, in the block format and as the restart-max-sec:
token:
service dialer {
restart = "always"
restart-sec = 10
restart-max-sec = 300
command = "dialer"
}
When set, the delay doubles for each restart, from restart-sec up to
the limit: 10, 20, 40, 80, 160, 300. Without restart-sec the first
restart is at once, then after 2, 4, 8 seconds. The delay follows the
restart counter, so it comes down again as the counter ages. With
restart always the counter stops where the delay peaks.
Fixes#306
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A crashing service with restart always waited five seconds before
every restart. It should wait two seconds for the first five, like a
service with the default restart-max:
service foo {
restart = "always"
command = "foo"
}
The back-off starts after half of restart-max attempts. For restart
always restart-max is -1, so half of it was zero and the back-off
began with the first restart. Use the default restart-max instead.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>