Proposed 1.0 semantics specification
cdahlin at redhat.com
Wed Jun 17 22:56:26 BST 2009
On 06/17/2009 05:33 PM, Scott James Remnant wrote:
> On Wed, 2009-06-17 at 15:20 -0400, Casey Dahlin wrote:
>>> This kind of "boot essential" job has been speculated for a while, but
>>> no design for it has come forwards.
>>> Upstart's design does inherently mean that services may simply not be
>>> started, and that's not considered an error condition.
>> The right way IMHO is to have a service/state called "system_normal"
>> or what have you, which depends on everything that should be running
>> every day. When that service is not running, there's a problem. If we
>> have cron functionality and inverted dependencies, we have everything
>> we need:
> That's certainly the simplest way to do it, though what would look to
> see whether that service is running or not (to report the problem) and
> how often would it check/how long after boot? :)
The way its worded in my example the check would happen 1 minute after
startup, and never again. An alternative would be a "settled" event,
which would occur every time the state machine stops transitioning for a
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the upstart-devel