Improve handling of OOM in CI
Categories
(Testing :: General, enhancement)
Tracking
(Not tracked)
People
(Reporter: jrmuizel, Unassigned)
References
Details
When we get an OOM it's pretty hard to make actionable, because we don't really have any information on what's actually causing the problem. Ideally, we'd have a memory report from the time of the crash, but that's probably not practical. It may still be valuable to have a report from a couple of minutes earlier because they leak could've happened then.
Another, idea could be to log memory usage before and after each test. That could give a hint to which tests were causing the most memory usage.
Comment 1•3 years ago
|
||
(In reply to Jeff Muizelaar [:jrmuizel] from comment #0)
Another, idea could be to log memory usage before and after each test. That could give a hint to which tests were causing the most memory usage.
We already do this part of it. The lines look like this:
MEMORY STAT | vsize 20983671MB | residentFast 1489MB
Updated•3 years ago
|
Comment 2•3 years ago
|
||
In browser-chrome, you can also scrape the ++DOMWINDOW --DOMWINDOW stuff to figure out how many windows during each teste I have a script somewhere that did that, though I did not get any useful conclusions from it. We have had issues in the distant past where the test harness was leaking windows.
Description
•