cron is simple, if its up the event fire if not, NOT.
So I think the behaviour is different to that what I expect from EJB
(and the spec), the event must at minimum fired once.
For the interval timer I would prefer the same behaviour as in AS5, it
will make the migration support of SEG easier ;)
For the calendar timer I'm not sure,
concurrent might not what the customer expect or hope.
And the idea for persistent c-timer might be that after start 'fire it
one time is enough'
If you have long period (e.g. once a day/week) I hope that there is
never be such long downtime that the timer is run twice after start;)
But customers might have the idea to schedule every x minutes in some
cases and then concurrent or catch up after downtime might not what is
wanted.
On 12/16/2011 03:36 PM, Jason Greene wrote:
Good questions. Perhaps we should compare with cron, quartz, etc
Sent from my iPhone
On Dec 16, 2011, at 8:23 AM, Wolf-Dieter Fink<wfink(a)redhat.com> wrote:
> I've tested the behaviour of AS7 timer during bug fixing.
> And now I need to discuss and clarify how the behaviour in AS7 should be.
>
>
> There is a big difference in case of overlapping timers between EAP5.1
> (and earlier, I don't check AS6) and AS7.
>
> ISSUE:
> If a timer is longer active as the interval, the time-out was delayed as
> long as the running timer is active.
> With AS7 the time-out method will be executed and run concurrent!
>
> I see migration issues because the application may rely on the non
> concurrent behaviour.
>
> QUESTIONS:
> The fix looks simple.
> But I want to clarify whether there is a counter-argument
>
> 1) recurring (interval) timer
> A interval timer should not be called until it is still running
> => there is no WARN/INFO should we add it?
>
> 2) calendar timer
> Should a calendar timer delayed also if the period between time-outs is
> to short?
> => I would prefer a WARN message if the timer is dropped
>
> What about catch up on startup if the timer is persistent.
> ATM all events are fired in parallel if there are more.
> (This will be a nice ERROR party in the log file in case of
> SingletonBeans with default WRITE lock.)
> => Drop it with a warning
> => block it until the first is finished
>
>
> regards
> Wolf
> _______________________________________________
> jboss-as7-dev mailing list
> jboss-as7-dev(a)lists.jboss.org
>
https://lists.jboss.org/mailman/listinfo/jboss-as7-dev