OpenRC, Upstart, systemd wishlist stuff
There seem to be a few new discussions about these possible solutions
As well as the traditional init scripts, I've worked with systemd on
Fedora and SMF on Solaris. Out of all possible solutions, I don't have
any strong feelings about which solution Debian should go with at this
stage.
However, one thing that comes to mind is the ease of integration with
management tools - as far as I know, Nagios users currently have to
a) manually declare service relationships (duplicating the LSB data from
their init scripts)
b) manually write action scripts that invoke init scripts or whatever
else is in use to try and restart a failed service.
In a more ideal environment (this is obviously a wishlist item, but an
important one), Nagios, NRPE or any similar tool would dynamically
discover the service inter-dependencies, mechanisms for checking state
and the corrective actions to use. The admin could then do little more
than setting some switch in Nagios that says `manage all my services
automatically'
To enable such convenient deployment, it is probably necessary for one
or more of these init solutions to expose sufficient data for automated
scanning.
Another nice wishlist item would be to have a single init/scheduler
paradigm that can work standalone or in a distributed manner, for load
balancing, HA and to ensure that services are only brought up when
services on other hosts are available. E.g. don't start Tomcat on host1
unless MySQL on host2 is up, or only run a particular cron job when the
necessary DB is up, or try to bring up the DB when the cron job wants to
run.
If jessie includes features like these then it would easily leapfrog the
advances that other distributions have made recently in progressing past
SysV init.
Reply to: