Thanks for those debug tips.
It doesn't seem to be a celery related problem.
Sessions are initiated after the fork and SQLAlchemy is not used as the
messaging backend (We use Redis).
Playing with the pool_recycle option in the Pyramid configuration I
clearly limited the number of exceptions.
On Wednesday, April 17, 2019 at 4:36:30 AM UTC-4, tonthon wrote:
>
> May the scoped_session factory, used in both services (web and celery),
> may cause such a problem ?
>
Apologies for my earlier reply, that would only affect exceptions in
Celery. I do suggest checking your Celery code to ma
Yes, celery is running in a separate service. The objects are never
passed from the web app to the celery service, only ids are passed,
Objects are then retrieved on the celery side.
May the scoped_session factory, used in both services (web and celery),
may cause such a problem ?
Le 16/04/
On Tuesday, April 16, 2019 at 9:41:39 AM UTC-4, Mike Bayer wrote:
>
> On Tue, Apr 16, 2019 at 4:49 AM tonthon wrote:
> >
> > Celery tasks are using the same scoped_session factory, could it cause
> the errors we're facing here ?
>
Is Celery running as threads or multiprocessing? IIRC, it u
Note also the objects can be copied using copy.deepcopy(), or pickle
loads()/dumps(), before sending them to the other Session. This
might not have worked in older Python / SQLAlchemy versions but
deepcopy() seems to rely upon the pickle internals e.g. __setstate__,
which in SQLAlchemy will cor
On Tue, Apr 16, 2019 at 4:49 AM tonthon wrote:
>
> Celery tasks are using the same scoped_session factory, could it cause the
> errors we're facing here ?
what's important is if these tasks run in the same process or not ? (
i thought celery runs as a separate service?) If you are passing
O
Celery tasks are using the same scoped_session factory, could it cause
the errors we're facing here ?
Le 15/04/2019 à 15:39, Mike Bayer a écrit :
On Mon, Apr 15, 2019 at 5:41 AM tonthon wrote:
I tried to set a lower value for the pool_recycle value and it seems to work.
There is a celery se
On Mon, Apr 15, 2019 at 5:41 AM tonthon wrote:
>
> I tried to set a lower value for the pool_recycle value and it seems to work.
>
> There is a celery service running in the background, maybe it could affect
> the session management.
this will reduce the problem but the architectural issue that
I tried to set a lower value for the pool_recycle value and it seems to
work.
There is a celery service running in the background, maybe it could
affect the session management.
Le 12/04/2019 à 15:58, Mike Bayer a écrit :
it's likely that a database connection is being returned to the pool
in
it's likely that a database connection is being returned to the pool
in an invalid state.
Switching to NullPool temporarily might reveal that this solves all
the issues ; at the very least, I would try setting pool_recycle to a
low number, like 5 minutes, however this won't prevent the problem,
ju
Le 10/04/2019 à 17:12, Mike Bayer a écrit :
On Wed, Apr 10, 2019 at 9:23 AM tonthon wrote:
Hi,
We're using sqlalchemy in a Pyramid Web Application.
We use the ZopeTransactionExtension and our session factory is initialized this
way :
DBSession = scoped_session(sessionmaker(extension=ZopeT
On Wed, Apr 10, 2019 at 9:23 AM tonthon wrote:
>
> Hi,
>
> We're using sqlalchemy in a Pyramid Web Application.
>
> We use the ZopeTransactionExtension and our session factory is initialized
> this way :
>
> >>> DBSession =
> >>> scoped_session(sessionmaker(extension=ZopeTransactionExtension()))
Hi,
We're using sqlalchemy in a Pyramid Web Application.
We use the ZopeTransactionExtension and our session factory is
initialized this way :
>>> DBSession =
scoped_session(sessionmaker(extension=ZopeTransactionExtension()))
We use the pyramid_tm that wraps each web requests in a transact
13 matches
Mail list logo
