Closed Bug 1484641 Opened 6 years ago Closed 6 years ago

Inactive bugs should be 'resolved incomplete' after the weekly intermittent report (orange factor) commented

Categories

(bugzilla.mozilla.org :: Bulk Bug Edit Requests, task)

Product:
bugzilla.mozilla.org
Component:
Bulk Bug Edit Requests
Version:
Production
Type:
task
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED FIXED

People

(Reporter: aryx, Unassigned)

References

Details

To prevent the bug from getting closed despite new failure occurrences during the last week.
Summary: Inactive bugs should be → Inactive bugs should be 'resolved incomplete' after the weekly intermittent report (orange factor) commented
The script that closes bugs is written/run by another team. Joel will know what component this belongs in.
Flags: needinfo?(jmaher)
I believe Emma runs this script
Flags: needinfo?(jmaher) → needinfo?(ehumphries)
At what time Sunday does the weekly update of failure occurrences run? It was my understanding that it runs at 5PM Pacific/Midnight UTC. I normally run the script to close bugs at 10PM Pacific, but it was run at 7PM last night.
Flags: needinfo?(emorley)
Flags: needinfo?(ehumphries)
Flags: needinfo?(aryx.bugmail)
Component: Intermittent Failures View → Bulk Bug Edit Requests
Product: Tree Management → bugzilla.mozilla.org
Version: --- → Production
It was bug #1482833 that changed the task back to Monday (also it ran last Monday late in the day, rather than Sunday).
I'll update my schedule to do that task after 7am UTC Mondays. Ideally this should be a scheduled batch job.
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → FIXED
I'm curious about this problem since previously the OrangeFactor commenter ran on Mondays around 1am UTC IIRC and with the replacement we moved it to Monday 7am UTC (although it ran much later last week). Why would bugs that received that comment last Monday still be closed Sunday night? What's the rule for closing bugs?

Sebastian, could you post a link to a bug where the commenter left a comment last Monday and the bug was still closed yesterday?

Emma, the potential issue with running your script right after the weekly commenter is that it can take quite a bit of time to run (hoping to optimize that at some point), but if it fails for any reason we'll receive an alert by 8am UTC and it'll need to be run manually (which means Ed would have to do it since I'm not awake then). But that assumes it was a database/query issue or something easily fixable, and then it takes time for it to run. I think it'd be better to have your job run first.
Flags: needinfo?(ehumphries)
Flags: needinfo?(aryx.bugmail)
The issue is that bugs get closed which have been used for classification during the last week but not the 2(?) weeks before.

Examples: bug 1451939, bug 1451939, bug 1464551, bug 1479190
Flags: needinfo?(aryx.bugmail)
Can we just combine the jobs? The query is simple, and we can set you up with a user who can close the bugs without sending bugspam?
Flags: needinfo?(ehumphries)
Sebastian, looking at those bugs, there's not a problem with the commenter. The commenter didn't run on Sunday Aug 5th, but even if it had there wouldn't have been failures posted for these bugs:

For the two weeks prior to today's weekly run, there were 0 failures:
https://treeherder.mozilla.org/api/failuresbybug/?startday=2018-07-29&endday=2018-08-12&tree=all&bug=1479190
https://treeherder.mozilla.org/api/failuresbybug/?startday=2018-07-29&endday=2018-08-12&tree=all&bug=1464551
https://treeherder.mozilla.org/api/failuresbybug/?startday=2018-07-29&endday=2018-08-12&tree=all&bug=1451939 

Emma, we could possibly add that functionality in, but at the moment we're not fetching bug metadata for every bug in the commenter - only ones that meet certain conditions. I'm not sure switching to a different user/api key as part of that task makes sense. Ed, what do you think? 

For now, could we move the bug closing task to Monday night or Tuesday am?
Flags: needinfo?(sclements) → needinfo?(emorley)
I can move the closing task to Tuesday my AM, and I'll look for an email from you or Ed if I need to hold off.
There is nothing wrong with the comments from the orangefactor/IFV commenter. It's only about the current situation that the OF/IFV bot now runs shortly after the bot which autocloses inactive bugs and due to that the autocloser lacks the update about the last 6-7 days to the bug, regards it as inactive and closes it.
There are multiple issues being discussed here:
1) When the bug commenter should run relative to the bug closing script (I agree the bug closing should occur after the commenter runs)
2) Regardless of order, what day of the week/rough time should the scripts should run
3) Automating the bug closing in some manner and whether this should remain as a separate script, or be moved into the commenter (pros and cons of the latter - as Sarah says, Treeherder doesn't process all the bugs, so implementation wouldn't overlap as much as one would think)
4) How we should coordinate the two scripts until #3 is done (ie: let Emma know if the commenter failed for whatever reason)

All issues apart from #3 seem to be resolved, and that issue is best discussed in a separate bug / mailing list post.
Flags: needinfo?(emorley)
To summarize:

1. Run the commenting script before the closing script.
2. The closing script should wait for the commenting script to complete its run.
  * For now, the commenting script runs on Mondays at 1am UTC
  * The closing script will run on Tuesdays at 9am Pacific (17 UTC)
3. TBD
4. If the script to comment bug fails or is delayed, then send :emceeaich an email notification to wait to run the closer script until further notice

Does that sound correct?
Yes, but currently the commenting script runs Mondays at 7am UTC :) Thanks Emma.
You need to log in before you can comment on or make changes to this bug.