Add option to restart Mopidy service

[jcass77]

It may be useful to have an option available to restart the Mopidy service to deal with situations where the actor dies unexpectedly (e.g. due to a misbehaving backend extension).

This would be quicker than having to do a full reboot, and easier than having to log in to the terminal to issue the command manually.

It probably means just adding WebServiceRestartRequestHandler that does service mopidy restart, similar to

mopidy-websettings/mopidy_websettings/__init__.py

Lines 143 to 151 in 1c55d94

	class WebRebootRequestHandler(tornado.web.RequestHandler):
	def initialize(self, core):
	self.core = core
	def post(self):
	logger.info('Halting system')
	os.system("sudo shutdown -r now")
	os.system("shutdown -r now")

?

[kingosticks]

I think when we played around before with having it restart itself it
didn't work. Presumably because when you stop you obviously then can't
start again! So you need to do the restart from another process. I think.
On 22 Nov 2015 04:37, "John Cass" notifications@github.com wrote:

It may be useful to have an option available to restart the Mopidy service
to deal with situations where the actor dies unexpectedly (e.g. due to a
misbehaving backend extension).

This would be quicker than having to do a full reboot, and easier than
having to log in to the terminal to issue the command manually.

It probably means just adding WebServiceRestartRequestHandler that does service
mopidy restart, similar to

mopidy-websettings/mopidy_websettings/__init__.py

Lines 143 to 151 in 1c55d94

	class WebRebootRequestHandler(tornado.web.RequestHandler):
	def initialize(self, core):
	self.core = core
	def post(self):
	logger.info('Halting system')
	os.system("sudo shutdown -r now")
	os.system("shutdown -r now")

?

—
Reply to this email directly or view it on GitHub
#17.

[kingosticks]

And in musicbox we have a process manager to restart mopidy (and other
things) if/when they crash. The name of it escapes me.
On 22 Nov 2015 13:11, "Nick Steel" kingosticks@gmail.com wrote:

I think when we played around before with having it restart itself it
didn't work. Presumably because when you stop you obviously then can't
start again! So you need to do the restart from another process. I think.
On 22 Nov 2015 04:37, "John Cass" notifications@github.com wrote:

It may be useful to have an option available to restart the Mopidy
service to deal with situations where the actor dies unexpectedly (e.g. due
to a misbehaving backend extension).

This would be quicker than having to do a full reboot, and easier than
having to log in to the terminal to issue the command manually.

It probably means just adding WebServiceRestartRequestHandler that does service
mopidy restart, similar to

mopidy-websettings/mopidy_websettings/__init__.py

Lines 143 to 151 in 1c55d94

	class WebRebootRequestHandler(tornado.web.RequestHandler):
	def initialize(self, core):
	self.core = core
	def post(self):
	logger.info('Halting system')
	os.system("sudo shutdown -r now")
	os.system("shutdown -r now")

?

—
Reply to this email directly or view it on GitHub
#17.

[jcass77]

Yes you may be thinking of monit.

The problem is that the Mopidy actor may be dead even though Mopidy itself keeps on responding on port 6680 (which is what monit checks for). This means that if an internal Mopidy error occurs that causes certain extensions to stop working monit will not always pick up on it.

I haven't been able to figure out how to get the service to restart itself though. Simply spawning a separate process to run the os command does not appear to be sufficient as the SIGTERM signal is propagated to all subprocesses, which means they all die when the main process goes down.

[kingosticks]

That is a good point.

You can probably send a command to monit and have it do the restart but that's not very helpful if you don't have monit running.

Simply spawning a separate process to run the os command does not appear to be sufficient as the SIGTERM signal is propagated to all subprocesses, which means they all die when the main process goes down.

I think it needs to be a different process group to stop this happening. Or maybe just use python-daemon?

[jcass77]

I'm not getting anywhere with this. I tried creating a temporary shell script that ignores SIGTERM and creates a subprocess from within the script:

script = '''#!/bin/sh
python -c 'import subprocess, signal\nsignal.signal(signal.SIGTERM, signal.SIG_IGN)\nsubprocess.Popen(["service", "mopidy", "restart"])'
'''
script_file = tempfile.NamedTemporaryFile('wt')
script_file.write(script)
script_file.flush()

subprocess.call(["sh", script_file.name], close_fds=True)

This, and everything else I've tried, just dies along with the main mopidy process.

Messing with the process groups and dealing with sessions and orphans seems pretty complicated (http://linux.die.net/man/2/setpgid).

Does anybody have any code examples that we could perhaps use as a baseline?