1838836 - Add TaskclusterMetadata to the alert summary API endpoint

Status: RESOLVED FIXED

(Tree Management :: Perfherder, enhancement)

Description

[Alexandru Ionescu (needinfo me) [:alexandrui]]

Add the taskcluster task_id and retry_id data to the alert summary API endpoint. This would help us getting the profile urls without fetching the entire jobList‎ in the frontend.
Also will help us being a step closer to getting the side by side task id directly in alerts view.
The field would go under alertsummary api ["results"]["alerts"][alerts_id] next to "series_signature".

Currently, we include the profile links to the alert summaries by copying the summary from alerts view, but they are not included in the comment 0 of the regression bug. If we want to do this we’ll run into a big performance issue that will make the browser hang on the alerts view.

This code is fetching the entire push list twice for every alert item that needs the profiler links. The performance difference is from few milliseconds when copying a small alert summary for talos, to thousands of milliseconds (around 4 seconds in the simulation in the screenshots) for browsertime - where we use profiler links. If we do this for a page of alerts when we load the alerts view, it will simply hang the browser. So including the data we need directly into the API response of the alert items.

Comment 1

[Alexandru Ionescu (needinfo me) [:alexandrui]]

Attachment: Screenshot from 2023-06-29 13-57-28.png

Comment 2

[Alexandru Ionescu (needinfo me) [:alexandrui]]

Attachment: Screenshot from 2023-06-29 13-58-19.png

Comment 3

[Acasandrei Beatrice (needinfo me)]

Have you done an investigation and concluded that the clipboard bug is related to the profiler link implementation done by Nazim?

"Currently, we include the profile links to the alert summaries by copying the summary from alerts view, but they are not included in the comment 0 of the regression bug. If we want to do this we’ll run into a big performance issue that will make the browser hang on the alerts view." Could you also provide more details regarding this statement? It's not clear what performance issue you are referring to. Doesn't comment 0 (in the bugs filed from Alerts View) contain the before/after links?

Comment 4

[Alexandru Ionescu (needinfo me) [:alexandrui]]

(In reply to Acasandrei Beatrice (needinfo me) from comment #3)

Have you done an investigation and concluded that the clipboard bug is related to the profiler link implementation done by Nazim?

The code that fetches the entire push list was there before, it wasn't implemented by Nazim. The clipboard bug comes from this code because it's an async code that fetches a big amount of data, and the clipboard code is problematic also. If you want I can double check to make sure this fetch code causes the issue.

"Currently, we include the profile links to the alert summaries by copying the summary from alerts view, but they are not included in the comment 0 of the regression bug. If we want to do this we’ll run into a big performance issue that will make the browser hang on the alerts view." Could you also provide more details regarding this statement? It's not clear what performance issue you are referring to. Doesn't comment 0 (in the bugs filed from Alerts View) contain the before/after links?

My bad, comment zero contains the before/after links (the bugs that I checked weren't containing before/after links), but there's a performance issue in the way the code fetches the push data to get the taskcluster task ids. If we want to make alerts view the canonical source of information, we can not use this code in alerts view (done something similar and had my code reverted twice due to browser hanging).