Rate Limits of Pull Request API endpoint?

I am building a check-script på a CI-env to check for new Pull Requests and run builds for them. The problem I am facing is that I do get alot of 429 errorcodes since I bombard the API to much (multiple repos and every X time-interval).

https://confluence.atlassian.com/bitbucket/rate-limits-668173227.html does not however document how many requests I am allowed to do for the endpoint(s):

/api/2.0/repositories/${project}/${repository}/pullrequests?limit=${limit}&state=OPEN

Also the document mentions that you are locked out for an hour which is not the way I "feel" how the API responds. But rather it works again just a few seconds later.

Is there any better way to do these checks and/or any info on how much I will have to throttle my requests?

1 answer

1 accepted

0 votes

Answer accepted

Hi @John Lindahl

Thank you for the detailed description of your scenario.

Let me share what I know about it:

The 1h limit is measured by a rolling window which means that in every minute you may have back some "credits" to run new requests until the total of 1000. That explains why you are not blocked for 1h.
Any request to /api/2.0/repositories/... will share the same request limit of 1000 calls.
It means that you call 500 times the /api/2.0/repositories/${project}/${repository}/pullrequest endpoint and also call 500 times the /api/2.0/repositories/{username} endpoint in less than 1h you will be blocked until the rolling window frees up some calls.
Each user will have a new set of limits. So you will have 1000 calls per hour for user A and also 1000 request per hour for user B. In case you have more users to run those calls this is how you could handle more calls per hour.

I hope I could clarify a little bit how the limits are set.

Please let me know if you have more questions.