Long-running process with many SQL operations crashes when DEBUG is True
|Reported by:||Patrick Paul||Owned by:||nobody|
|Component:||Database layer (models, ORM)||Version:||1.5|
|Has patch:||no||Needs documentation:||no|
|Needs tests:||no||Patch needs improvement:||no|
When DEBUG=True, SQL queries are buffered in memory and for long-running processes this can cause the server to kill the process.
In my particular use case, I was crashing when executing a South data migration which ingested large datasets into my PostGIS database.
There is a monkey patch available on SO: http://stackoverflow.com/a/7769117/1545769
After including this monkey patch in each of my data migration files, the migrations completed successfully.
Here is why I am creating an issue:
1) There isn't a very helpful message when the process is terminated; you have to deduce it from looking at PostgreSQL logs or /var/log/syslog to see that low memory was the perpetrator.
2) There is no way to configure in settings.py to disable the CursorDebugWrapper (e.g. settings.DISABLE_DEBUG_CURSOR boolean)
3) I wonder if we can't make Django smart enough to trap Out-of-memory errors to present an informative exception via Python and/or...
4) ...smartly manage the connection.queries buffer to prevent it from using too much RAM in the first place?
N2 would be an easy pickings solution (here: https://github.com/django/django/blob/master/django/db/backends/__init__.py), but it wouldn't solve N1 and developers might not know that RAM is the issue. So I think N3/N4 should be considered.