Re: [google-appengine] Re: Chaining tasks forever and checking enqueued tasks

2010-11-21 Thread Ian Lewis
On Mon, Nov 22, 2010 at 9:14 AM, Nick Johnson (Google) < nick.john...@google.com> wrote: > Hi Ian, > > On Mon, Nov 22, 2010 at 11:10 AM, Ian Lewis wrote: > >> While the task name thing should generally work, they don't provide a >> guarantee that inserting the task >> won't succeed more than once

Re: [google-appengine] Re: Chaining tasks forever and checking enqueued tasks

2010-11-21 Thread Nick Johnson (Google)
Hi Ian, On Mon, Nov 22, 2010 at 11:10 AM, Ian Lewis wrote: > While the task name thing should generally work, they don't provide a > guarantee that inserting the task > won't succeed more than once. > A named task will never be duplicated - that's the point of task names - though any task at al

Re: [google-appengine] Re: Chaining tasks forever and checking enqueued tasks

2010-11-21 Thread Ian Lewis
While the task name thing should generally work, they don't provide a guarantee that inserting the task won't succeed more than once. Also, I don't particularly want to have to add a task to check if it is inserted. There is also the problem of tombstoning names, so I would need to come up with uni

[google-appengine] Re: Chaining tasks forever and checking enqueued tasks

2010-11-21 Thread vlad
Agree with Robert named tasks are the way to go. I have a similar requirements - have a long running state machine implemented as chained tasks. Once I worked out transactional issues and added "choke" points to prevent fork bomb it has been running ok for over 1 week now. Clarification, by "runnin

[google-appengine] Re: Chaining tasks forever and checking enqueued tasks

2010-11-21 Thread Remigius
Ian, It does not look like there's a facility to look up queued tasks in the task queue API. Instead you could use a heartbeat in the data store (i.e. save an entity with a timestamp each time the task queue task runs) that is checked by the cron job. Cheers, Remigius. -- You received this mess