you know that on console, after queue_task() you need to commit(), right ?
If you leave the transaction hanging, no wonder that you're observing
deadlocks!
On Friday, February 27, 2015 at 3:38:19 AM UTC+1, Osman Masood wrote:
>
> I was actually able to easily reproduce this. Just open up 2 web2py
> consoles (which connect to the same DB), and do:
>
> scheduler.queue_task('myfunc', pvars=dict(), immediate=True)
>
> The first one will succeed, and the second one will hang there for some
> time, and then give the deadlock exception above.
>
> This kinda makes sense, looking at the above DB query & considering
> record-level locking. We have conditional writes on all records in the
> scheduler_worker table, so if two connections attempt this at the same
> time, it would follow that there is a deadlock. (Correct me if I'm wrong.)
> So I think the solution would be something like (on line 1303 of
> scheduler.py):
>
> if immediate:
>
> - self.db(self.db.scheduler_worker.is_ticker ==
> True).update(status=PICK)
>
> + scheduler_worker_tickers =
> self.db(self.db.scheduler_worker.is_ticker == True).select()
>
> + for scheduler_worker_ticker in scheduler_worker_tickers:
>
> + scheduler_worker_ticker.update_record(status=PICK)
>
>
> On Thursday, February 26, 2015 at 3:54:47 PM UTC-8, Osman Masood wrote:
>>
>> Thanks for the quick response!
>>
>> I'm using Amazon RDS (just basically MySQL), on a pretty beefy instance
>> (db.m3.large), 1000 IOPS. Would be pretty surprised if that were the
>> problem.
>>
>> Running 5 scheduler workers with the default heartbeat (3s).
>>
>> On Thursday, February 26, 2015 at 3:35:41 PM UTC-8, Niphlod wrote:
>>>
>>> it seems to me that there is database contention........ although the
>>> exception **should** be trapped (i.e. try to set the status to "PICK", if
>>> not, well.... it's not a fatal error), there's something wrong with your
>>> setup or your database is too much "underpowered" to do what you're asking.
>>> How many workers are running and with what heartbeat ?
>>>
>>> On Friday, February 27, 2015 at 12:28:28 AM UTC+1, Osman Masood wrote:
>>>>
>>>> Hello all,
>>>>
>>>> I keep getting this error when calling scheduler.queue_task() with
>>>> immediate=True:
>>>>
>>>> <class 'gluon.contrib.pymysql.err.InternalError'> (1213, u'Deadlock
>>>> found when trying to get lock; try restarting transaction')
>>>>
>>>> The stack trace shows this is the culprit:
>>>>
>>>> self.db(self.db.scheduler_worker.is_ticker == True).update(status=PICK)
>>>>
>>>> (This is a web application which uses the scheduler pretty extensively.)
>>>> Seems like multiple connections are competing for access to the same
>>>> scheduler_worker row.
>>>>
>>>> Seems to me like a web2py bug. Any help would be greatly appreciated.
>>>> Thanks!
>>>>
>>>>
--
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 [email protected].
For more options, visit https://groups.google.com/d/optout.
