In one of my management scripts (which runs continuously, after setting up
a web2py environment), I copy a complete sqlite database directory from
another server (copied_db.sqlite, and *.table), open them with
"DAL('sqlite://copiedfile.sqlite', auto_import=True, path='/tmp/copy_path)".
I copy the files back and forth, and recreate the DAL object every 10-20
seconds or so. Note that on linux, this means that the sqlite file opened
every time is new (new copy replaces old one), even if it has the same name
and same contents as before -- this use case is different than standard
web2py use, in which the DAL always refers to the same file on disk (same
inode, same everything).
I noticed that after a few hours, memory consumption becomes huge (goes
from 7M to 400M in less than an hour, _and_ continues to grow, despite the
files not getting larger). At some point, the sqlite dbapi connect starts
failing as well.
Is there a pool or cache that keeps DAL objects alive behind the scenes?
I have so far been unable to find what would keep the DALs alive; where
should I look? The culprit could be anything that indirectly keeps a
reference to the connection or the database, e.g. - a Field, a Table,
anything like that. Any ideas where to look in web2py? my code doesn't
store any of these objects, by maybe something in web2py does?
