hum...how can cat3 be queued if cat3 depends on cat2 being executed ?
Or what you wanted to know is how to give a higher priority to cat1 tasks
instead of cat2 tasks (i.e. they are not related, it's just that you prefer
cat2 remaining in the queue a bit longer if there are cat1 tasks)?
In that case, you should organize tasks into groups (i.e. cat1 with
"high_prio", cat2 with "normal_prio", cat3 with "low_prio"), and then start
the workers with group_names.
Although the scenario for such things requires a bit of workers (i.e.
10-15) and to start 7 workers with high_prio, 2 with "normal_prio" and 1
with "low_prio"
i.e. -K
appname:high_prio,appname:high_prio,appname:high_prio,appname:high_prio,appname:high_prio,appname:high_prio,appname:high_prio,appname:normal_prio,appname:normal_prio,appname:low_prio
and that you can make the scheduler know that it should assign more tasks
of high_prio than normal_prio, but with a lower number of workers e.g.
-K
appname:high_prio:high_prio:high_prio:normal_prio:normal_prio:low_prio,appname:high_prio
you'd get what I call a "proportional dispatch"meaning it would
obviously "slim" the task queue of "high_prio" before "low_prio" just
giving more chances to high_prio tasks to be executed vs low_prio. All this
works best if you are in the situation of having a long queue of tasks
ready to be processed ( start_time <= request.now).
If you instead want to prioritize "scientificly" a little known fact (but
it's shown in the scheduler test app and there's a regression test for it)
is that tasks are picked up ordered by next_start_time. So, again, if you
are queuing tasks that should be processed ASAP, just queue high_prio (your
"cat1") with next_run_time = request.now - datetime.timedelta(days=90),
normal_prio with next_run_time=request.now - datetime.timedelta(days=60)
and low_prio with request.now.
On Wednesday, October 26, 2016 at 4:01:07 PM UTC+2, Pierre wrote:
>
> yes i am aware of the scheduler new deps feature
> this would require to know every single task id to build the parent
> children relation
> I don't know what's in the queue, tasks are queued in arbitrary order and
> the queue evolves constantly so i can't scan single task ids
>
> alltogether the scheduler does a pretty good job
>
> I encounter this only once :
>
> *2016-10-25 14:55:28,137 - web2py.scheduler.hp533#4326 - ERROR - Error
> coordinating TICKER *
>
> is this some kind of fatal error ?
>
>
>
--
Resources:
- http://web2py.com
- http://web2py.com/book (Documentation)
- http://github.com/web2py/web2py (Source code)
- https://code.google.com/p/web2py/issues/list (Report Issues)
---
You received this message because you are subscribed to the Google Groups
"web2py-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to web2py+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
