Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
-
Github
-
Airflow
Description
We have a problem that pretty much all PRs in Apache Airflow project get peridically updated, which prevents us from closing them automatically by stalebot.
I already saw it several times - it happens from time to time (every few weeks) that all PRs in Apache Airflow are somehow "refreshed" - i.e. their updated date changes to "current" even if there is no change visible for those PRs. It's quite clearly visible for example here: https://github.com/apache/airflow/pulls?page=8&q=is%3Apr+is%3Aopen+sort%3Aupdated-desc - you can see clearly that pretty much all the PR - including some really old ones - were updated yesterday. But when you look at those issues you will see that they were last updated months ago and there is no activity from yesterday.
For example look at this one: https://github.com/apache/airflow/pull/2460 - it's been last updated in December 2018 (!) and still when you sort tasks by 'most recently updated' it is shown as updated yesterday.
This "update date" gets bumped periodically. I opened ticket to Github so that they could check what happens but their response was not helpful (see below).
I looked at other apache repositories and they seem to be free of this issue - for example here: https://github.com/apache/lucene-solr/pulls?page=5&q=is%3Apr+is%3Aopen+sort%3Aupdated-desc - you would see rightfully some very old PRs and no PRs are updated at the same time.
I was wondering that maybe there is some write-capable integration for Apache Airflow project installed that performs such generic "refresh" - I have no access to settings of the repo, but maybe you can find the possible culprit for this behaviour.
Here is the response from Github I got:
##############################################################
Our apologies for the wait, and thank you for getting in touch! Due to a high volume of requests, we are currently experiencing much longer than average response times here in Support. You asked:
Can you please let us know what action caused the update and what can we do to prevent it from happening again ?
The updated_at for any object, including users, will change whenever the database record for that object is updated. Such database updates can happen for many reasons, though we don't have a complete list of those to share with you and your team. We wish could be of more help here as we see how this can be a problem for you and your team, but we don't currently have any other insight to share at this time.
Please let us know how else we can be of help!
##########################################################
I already saw it several times - it happens from time to time (every few weeks) that all PRs in Apache Airflow are somehow "refreshed" - i.e. their updated date changes to "current" even if there is no change visible for those PRs. It's quite clearly visible for example here: https://github.com/apache/airflow/pulls?page=8&q=is%3Apr+is%3Aopen+sort%3Aupdated-desc - you can see clearly that pretty much all the PR - including some really old ones - were updated yesterday. But when you look at those issues you will see that they were last updated months ago and there is no activity from yesterday.
For example look at this one: https://github.com/apache/airflow/pull/2460 - it's been last updated in December 2018 (!) and still when you sort tasks by 'most recently updated' it is shown as updated yesterday.
This "update date" gets bumped periodically. I opened ticket to Github so that they could check what happens but their response was not helpful (see below).
I looked at other apache repositories and they seem to be free of this issue - for example here: https://github.com/apache/lucene-solr/pulls?page=5&q=is%3Apr+is%3Aopen+sort%3Aupdated-desc - you would see rightfully some very old PRs and no PRs are updated at the same time.
I was wondering that maybe there is some write-capable integration for Apache Airflow project installed that performs such generic "refresh" - I have no access to settings of the repo, but maybe you can find the possible culprit for this behaviour.
Here is the response from Github I got:
##############################################################
Our apologies for the wait, and thank you for getting in touch! Due to a high volume of requests, we are currently experiencing much longer than average response times here in Support. You asked:
Can you please let us know what action caused the update and what can we do to prevent it from happening again ?
The updated_at for any object, including users, will change whenever the database record for that object is updated. Such database updates can happen for many reasons, though we don't have a complete list of those to share with you and your team. We wish could be of more help here as we see how this can be a problem for you and your team, but we don't currently have any other insight to share at this time.
Please let us know how else we can be of help!
##########################################################