Add support for distributed run #77
Conversation
When tests are distributed, a copy of the database is created for each worker at the start of the test session. This is useful to avoid concurrent access to the same database, which can lead to deadlocks. The provided database is therefore used only as template. At the end of the tests, all the created databases are dropped.
lmignon
force-pushed
the
master-distributed-run
branch
from
4e02c7f
to
9ee223f
Compare
|
Codecov ReportAttention: Patch coverage is
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## master #77 +/- ##
=========================================
Coverage ? 33.65%
=========================================
Files ? 1
Lines ? 104
Branches ? 0
=========================================
Hits ? 35
Misses ? 69
Partials ? 0 ☔ View full report in Codecov by Sentry. |
|
ping @yvaucher |
pytest_odoo.py
Outdated
| ret = os.system(f"psql -lqt | cut -d \| -f 1 | grep -w {db_name}") | ||
| if ret == 0: | ||
| os.system(f"dropdb {db_name}") | ||
| os.system(f"createdb -T {original_db_name} {db_name}") |
There was a problem hiding this comment.
this sounds a very opinionated way to connect to the database, probably we should at least add such limitation in the Readme file or read odoo config file to retrieves host and credentials informations
There was a problem hiding this comment.
You can already configure psql with env var to get to the right config with PGDATABASE, PGUSER, PGHOST and so on. Using psql, dropdb and createdb without params seems fine to me.
|
@lmignon should we add a note on Did you test it? It groups tests by class to send related tests to the same worker (thus I would assume on the same DB), it might be a bit for efficient with Savepoints? https://pytest-xdist.readthedocs.io/en/stable/changelog.html#id189 |
Co-authored-by: Yannick Payot <[email protected]>
lmignon
force-pushed
the
master-distributed-run
branch
from
320b456
to
0168a0e
Compare
@yvaucher I've just tested it following your comment. I didn't notice any significant difference between the different ways of distributing tests between workers. |
|
Played a bit with it, and works fine, let's merge this. I'll create a small release to make it available. |
When tests are distributed, a copy of the database is created for each worker at the start of the test session. This is useful to avoid concurrent access to the same database, which can lead to deadlocks. The provided database is therefore used only as template. At the end of the tests, all the created databases are dropped.