improvement on runit scripts

[juruen@…]

We use runit to start/stop eBox services like: squid, samba, ntp... We are doing a very primitive use of runit at the moment. Basically, we just let runit start the services everytime they go down by unknown reasons.

This behaviour can be dramatically improved and we should at least do the following:

• Log whenever a service has been restarted
• Do not try to restart a service infinitely, that is, if there's an error and a service cannot be started, runit will be trying to start the service all the time. A better approach would be to set a period of time, and a limit of restarting attempts during that period, in case of reach that limit it should stop trying and log the issue properly.
• The log will be watched by EBox::Event::Watcher::Runit to inform administrator what it happens

[ejhernandez@…]

I will do it right now!

Event watcher to that problem is added

[ejhernandez@…]

(In [7785]) First runit wrapper script version refs #561

[ejhernandez@…]

(In [7796]) * Added script to be called in finish scripts inside the runit arch to log how many times a service has been restarted in a time interval. If maximum restart number is exceeded, the eBox module which owns the service will be stopped

• Added a configuration script to set the maximum number of restarts and the time interval
• apache-perl in apache eBox module is already controlled by this new framework

refs #561

[ejhernandez@…]

(In [7797]) Transform those files within tools/runit/ subdir which end at .finish to be the finish scripts in the corresponding runit directory refs #561

[ejhernandez@…]

(In [7798]) Return false when the service managed by runit is executing finish script refs #561

[ejhernandez@…]

(In [7799]) * Added Runit event watcher to check if some service is out of control to notice the eBox admin in some way

• Change return value from run to get back an array ref of EBox::Event instead a single EBox::Event since Runit watcher may return more than one event at the same execution

refs #561

[ejhernandez@…]

(In [7800]) * Updated pipe which communicates watchers and dispatchers to flush every new entry

• Adapting to the new return value from EBox::Event::Watcher::run method [7799]

closes #561 (spent 11.5)