By default, Airflow only allows creating a maximum of 1024 dynamic tasks at a time (this is configurable but it's not advised to raise this number). The alter_data step of the data refresh fails in the image data refresh because it tries to create batches of 100_000 records at time, but there are so many image records to process that we exceed the number of available batches (we get about 9k batches) and overload XCOMs.
This PR updates the alter_data steps to allow the batch size to be configured by media type, and updates image to use batches of 1,000,000 records. It still respects the configured DATA_REFRESH_ALTER_BATCH_SIZE environment variable as well.
Testing Instructions
Run the staging_image_data_refresh locally. You'll see that it creates batches of 1000 records; this is because we have DATA_REFRESH_ALTER_BATCH_SIZE configured to 1k locally.
Now comment that variable out in your catalog/.env. Because of #5099 you will also need to comment it out in catalog/env.template to prevent it from being added right back in. Now run ov j down -v && ov j up. Run the DAG again and you will see that the alter_data step puts all the records in one batch (because we have <1M records locally).
Checklist
[ ] My pull request has a descriptive title (not a vague title likeUpdate index.md).
[ ] My pull request targets the default branch of the repository (main) or a parent feature branch.
[ ] My commit messages follow [best practices][best_practices].
[ ] My code follows the established code style of the repository.
[ ] I added or updated tests for the changes I made (if applicable).
[ ] I added or updated documentation (if applicable).
[ ] I tried running the project locally and verified that there are no visible errors.
[ ] I ran the DAG documentation generator (ov just catalog/generate-docs for catalog
PRs) or the media properties generator (ov just catalog/generate-docs media-props
for the catalog or ov just api/generate-docs for the API) where applicable.
Developer Certificate of Origin
```
Developer Certificate of Origin
Version 1.1
Copyright (C) 2004, 2006 The Linux Foundation and its contributors.
1 Letterman Drive
Suite D4700
San Francisco, CA, 94129
Everyone is permitted to copy and distribute verbatim copies of this
license document, but changing it is not allowed.
Developer's Certificate of Origin 1.1
By making a contribution to this project, I certify that:
(a) The contribution was created in whole or in part by me and I
have the right to submit it under the open source license
indicated in the file; or
(b) The contribution is based upon previous work that, to the best
of my knowledge, is covered under an appropriate open source
license and I have the right under that license to submit that
work with modifications, whether created in whole or in part
by me, under the same open source license (unless I am
permitted to submit under a different license), as indicated
in the file; or
(c) The contribution was provided directly to me by some other
person who certified (a), (b) or (c) and I have not modified
it.
(d) I understand and agree that this project and the contribution
are public and that a record of the contribution (including all
personal information I submit with it, including my sign-off) is
maintained indefinitely and may be redistributed consistent with
this project or the open source license(s) involved.
```
Fixes
Description
By default, Airflow only allows creating a maximum of 1024 dynamic tasks at a time (this is configurable but it's not advised to raise this number). The
alter_datastep of the data refresh fails in the image data refresh because it tries to create batches of 100_000 records at time, but there are so many image records to process that we exceed the number of available batches (we get about 9k batches) and overload XCOMs.This PR updates the
alter_datasteps to allow the batch size to be configured by media type, and updates image to use batches of 1,000,000 records. It still respects the configuredDATA_REFRESH_ALTER_BATCH_SIZEenvironment variable as well.Testing Instructions
Run the
staging_image_data_refreshlocally. You'll see that it creates batches of 1000 records; this is because we haveDATA_REFRESH_ALTER_BATCH_SIZEconfigured to 1k locally.Now comment that variable out in your
catalog/.env. Because of #5099 you will also need to comment it out incatalog/env.templateto prevent it from being added right back in. Now runov j down -v && ov j up. Run the DAG again and you will see that thealter_datastep puts all the records in one batch (because we have <1M records locally).Checklist
Update index.md).main) or a parent feature branch.ov just catalog/generate-docsfor catalog PRs) or the media properties generator (ov just catalog/generate-docs media-propsfor the catalog orov just api/generate-docsfor the API) where applicable.[best_practices]: https://git-scm.com/book/en/v2/Distributed-Git-Contributing-to-a-Project#_commit_guidelines
Developer Certificate of Origin
Developer Certificate of Origin
``` Developer Certificate of Origin Version 1.1 Copyright (C) 2004, 2006 The Linux Foundation and its contributors. 1 Letterman Drive Suite D4700 San Francisco, CA, 94129 Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. Developer's Certificate of Origin 1.1 By making a contribution to this project, I certify that: (a) The contribution was created in whole or in part by me and I have the right to submit it under the open source license indicated in the file; or (b) The contribution is based upon previous work that, to the best of my knowledge, is covered under an appropriate open source license and I have the right under that license to submit that work with modifications, whether created in whole or in part by me, under the same open source license (unless I am permitted to submit under a different license), as indicated in the file; or (c) The contribution was provided directly to me by some other person who certified (a), (b) or (c) and I have not modified it. (d) I understand and agree that this project and the contribution are public and that a record of the contribution (including all personal information I submit with it, including my sign-off) is maintained indefinitely and may be redistributed consistent with this project or the open source license(s) involved. ```