Store longer unittest timelines in ActiveData

Assigned to


5 months ago
5 months ago


(Reporter: ekyle, Assigned: ekyle)


Firefox Tracking Flags

(Not tracked)



(1 attachment)



5 months ago
Created attachment 8853013 [details]

ActiveData is full, and only able to store 3 or 4 weeks of unittest data.

There are some solutions to this problem:

1. slim down the denormalized records to free up space
2. spend more money for larger cluster
3. stop running near-a-billion tests a week

We will attempt to act on #1.  Please help identify properties we should never need.

a. An example record is attached.
b. An old overview of the properties
do we collect information from the try branch?  Possibly that would be a way to greatly reduce the load here- maybe there are use cases related to try which make sense to keep the data for, it would be good to figure out what branches we monitor and what value we get from each branch.

Comment 2

5 months ago
Try accounts for 38% of the records.
I would prefer to not have any data from try. All sorts of things happen on try that never get pushed to trunk, and we can't easily relate events on try to events on other trees.

Comment 4

5 months ago
Removing try, removing most of the `task.*` properties, finding other bugs.
You need to log in before you can comment on or make changes to this bug.