feat: backfill data to clickhouse in batches and same thread #74
Conversation
openedx-webhooks
added
the
open-source-contribution
|
Thanks for the pull request, @Ian2012! Please note that it may take us up to several weeks or months to complete a review and merge your PR. Feel free to add as much of the following information to the ticket as you can:
All technical communication about the code itself will be done via the GitHub pull request interface. As a reminder, our process documentation is here. Please let us know once your PR is ready for our review and all tests are green. |
Ian2012
force-pushed
the
cag/add-batching-for-backfill
branch
from
d8d4c80 to
bb364df
Compare
event_sink_clickhouse/management/commands/dump_data_to_clickhouse.py
Outdated
Show resolved
Hide resolved
event_sink_clickhouse/management/commands/dump_data_to_clickhouse.py
Outdated
Show resolved
Hide resolved
| """ | ||
| Return the queryset to be used for the insert | ||
| """ | ||
| return self.get_model().objects.all() | ||
| if start_pk: | ||
| return self.get_model().objects.filter(pk__gt=start_pk).order_by("pk") |
There was a problem hiding this comment.
One last performance improvement we could make here is to use .values() or .values_list(), which would save a bunch of overhead on creating model instances, instead returning a dict or tuple. values_list performs best, but dealing with tuples can be annoying. We could also then choose the fields we want to select to just what we use, potentially saving a lot of data transfer from the db and memory on the job.
There was a problem hiding this comment.
Dealing with values_list would break the serializer, while values would return the data already partially serialized (still needs to deal with dates and related names). and we would need to implement a method serializer for every sink. Not sure about the gains here
There was a problem hiding this comment.
Yeah, I've seen some pretty significant gains in the past from it, but not always. If you're comfortable with the performance as it is we can leave it and revisit later if necessary.
Ian2012
force-pushed
the
cag/add-batching-for-backfill
branch
2 times, most recently
from
1c5281e to
e752256
Compare
itsjeyd
added
the
core contributor
Ian2012
force-pushed
the
cag/add-batching-for-backfill
branch
from
b2d450d to
358b2cf
Compare
fix: query course overviews with django model fix: remove submitted_objects objects chore: quality fixes chore: remove old dump courses to clickhouse command fix: fix initial page, tests, and quality chore: handle pr comments test: improve coverage fix: paginator last page is included test: fixing test
test: add test for get_queryset test: add test for get_queryset
Ian2012
force-pushed
the
cag/add-batching-for-backfill
branch
from
f8d341d to
a43189d
Compare
Ian2012
force-pushed
the
cag/add-batching-for-backfill
branch
from
a43189d to
755f961
Compare
test: add test for course_published changes chore: quality fixes test: add test for get_queryset changes in sinks test: add test for should_dump_item in course_published chore: use invalid emails chore: quality fixes test: add tests for course_published should_dump_item tmp
Ian2012
force-pushed
the
cag/add-batching-for-backfill
branch
from
84d9c64 to
69fa84e
Compare
|
@Ian2012 🎉 Your pull request was merged! Please take a moment to answer a two question survey so we can improve your experience in the future. |
Description
This PR improves the backfill command to send data in configurable batches and dismisses the use of Celery. This doesn't affect the usual LMS traffic and backfill time is reduced because clickhouse likes batch inserts and serialization is more performant.
This PR:
select_relatedwhen needed.