1046150 - Eideticker should use a real database of some kind for test data

Status: RESOLVED WONTFIX

(Testing Graveyard :: Eideticker, defect)

Description

[William Lachance (:wlach)]

Right now we're just storing the results of our test runs as static json. This doesn't really scale and causes weird issues like bug 1046066. Options:

1. Use / extend datazilla / treeherder.
2. Create a small web application that provides a web interface to a mysql database.
3. Write data directly to an elasticsearch database.

I think we decided that (1) wouldn't work very well for eideticker's data model (especially the out of band stuff: bug 909796). I was thinking (2) for a while but after talking with :ekyle yesterday I'm now thinking that just dumping the data to elasticsearch might be the way to go. Its unstructured data schema would be perfect for storing and querying eideticker's extra metadata (video url, per-frame entropy values,  ...). Also, it would remove the need for Kyle to maintain a script to periodically fetch eideticker data to put it into elasticsearch. Finally, it should be easy for other people to plug into the elastic search database (or a clone of it) and run queries against it. Need to investigate further.

Comment 1

[William Lachance (:wlach)]

I have been working on this. I have the views part basically done, just need to import the data and the dashboard. In the end, I decided to go for a sqlite3 database using django, as I think it'll be less trouble to manage than any of the alternatives. People can pull data from it to convert to elasticsearch as needed (in fact i should be able to add queries/views that make this easier than it was before).

Comment 2

[BMO Automation]

Eideticker has been discontinued, see bug 1361056