Improve handling of Jira's retry-after handling #1825
Merged
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
The time Jira sends in the Retry-After header is the minimum time Jira wants us to wait before retrying our request. However, the former implementation used this as a maximum waiting time for the next request. In result, there was a chance that we reached three retries without reaching the time that Jira expected us to wait and our request would fail. This implementation does also affect the other retry cases, as while previously we jittered our backoff between 0 and the target backoff, we now only jitter between 50% and 100% of the target backoff. However, this should still protect us from thundering herds and safes us from introducing a new minimum backoff variable for the retry-after case. This solves one of the issues reported in pycontribs#1805.
When rejecting request with a 429 response, Jira sometimes sends a Retry-after header asking for a backoff of 0 seconds. With the existing retry logic this would mark the request as non-retryable and thus fail the request. With this change, such requests are treated as if Jira had send a retry-after value of 1 second. This solves one of the issues reported in pycontribs#1805.
matthias-bach-by
had a problem deploying
to
ack
GitHub Actions
Failure
matthias-bach-by
had a problem deploying
to
cloud
GitHub Actions
Failure
|
The cloud tests are failing in set-up and this seems to be a wider issue unrelated to my changes as I am also seeing that on other PRs. Happy to rebase once that is fixed on main. |
adehad
approved these changes
Mar 21, 2024
4 tasks
|
Thanks for the excellent PR and sorry for the delay in getting in merged and released! |
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.
As reported in #1805, while 3.6.0 brought a beneficial change in evaluating Jira's retry-after header, in some scenarios it actually degraded the success rate of requests when hitting rate-limits due to two issues:
To address this, this PR changes the back-off behaviour to treat the retry-after time as a minimum backoff, at least as long as our maximum back-off time does allow this, and clips the retry-after header, if specified, to a minimum of one second.
I have extended the existing tests for the retry logic to assert this behaviour. If you would prefer a separate set of tests or see other changes to the approach, let me know and I'll adjust the PR.