feat: 🎸 replace Queue.add_job with Queue.upsert_job #694
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
upsert_job ensures there is only one waiting job for the same set of parameters. On every call to upsert_job, all the previous waiting jobs for the same set of parameters are cancelled, and a new one is created, with a new "created_at" date, which means it will be put at the end of the queue. It will help to fight against datasets that are updated very often (eg every minute): they will be treated only when there is available workers.
|
See for example: https://huggingface.co/datasets/atokforps/latent_v1_fullrun_alpha3_13/commits/main. Two commits are pushed every minute. The queues contain thousands of waiting jobs.
|
|
The issue is effectively fixed.
For the "/first-rows" queue, we will have first to wait for the /splits to be run again on them |
|
OK, #695 helped reduce the number of waiting jobs for /first-rows. There is no way to go further, since there are no more duplicates (for example, allenai/nllb has 2.656 splits, thus up to 2.656 jobs) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
upsert_job ensures only one waiting job for the same set of parameters. On every call to upsert_job, all the previous waiting jobs for the same set of parameters are canceled, and a new one is created with a new "created_at" date, which means it will be put at the end of the queue. It will help to fight against datasets that are updated very often (eg, every minute): they will be treated only when there are available workers.
It's a quick PR to fix the issue that the queues are increasing faster than the availability of the workers and that most of these jobs will be later skipped. Better to reduce the number of results in the queries (by reducing the number of waiting jobs).