[
https://issues.apache.org/jira/browse/NIFI-14355?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17934306#comment-17934306
]
Pierre Villard commented on NIFI-14355:
---------------------------------------
I think changing the default to something like 5 minutes could be a good first
step. But I think that for very large NiFi users that would not be enough to
not hit the rate limit. Then the question is finding the right balance between
rate limits and user experience... It is nice to not wait for a long time to
see that a new version is available. Maybe another approach would be to set the
background thread frequency to something much higher but we would check for new
versions when someone in the UI is doing a refresh - not sure if we distinguish
the normal background refresh and a forced refresh by someone in the UI. But
even in this case not sure if that's great because I find myself spamming the
refresh quite a lot :)
> Authenticated GitHub Registry Client hitting rate limit (5000 reqs per hour)
> ----------------------------------------------------------------------------
>
> Key: NIFI-14355
> URL: https://issues.apache.org/jira/browse/NIFI-14355
> Project: Apache NiFi
> Issue Type: Improvement
> Components: Flow Versioning, NiFi Registry
> Affects Versions: 2.2.0
> Environment: Not relevant
> Reporter: Andrea Di Stefano
> Priority: Major
> Attachments: image-2025-03-11-16-47-39-522.png
>
>
> Hello.
> We are using the GitHub Registry Client with a Personal Authentication Token
> (PAT) and constantly hitting the API rate limit. The rate limit, [as stated
> in the GitHub
> docs|https://docs.github.com/en/rest/using-the-rest-api/rate-limits-for-the-rest-api?apiVersion=2022-11-28#primary-rate-limit-for-authenticated-users],
> is 5000 requests per hour.
> We have a pipeline with 53 instances of 3 versioned process groups, which are
> part of a bigger pipeline (~1.1k components).
> I uploaded a related stack trace [here|https://pastebin.com/R0tTw4bP].
> Calling the GitHub API with Postman, we saw that the response headers include
> data about the rate limit, current usage, and reset time in Unix epoch
> format, see the attached screenshot for an example:
> !image-2025-03-11-16-47-39-522.png!
> Is NiFi periodically polling the GitHub API in background?
> Can we somehow limit the number of calls?
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
