1046060 - fxOS: Run Flame perf tests with 512MB configuration instead of auto (1GB)

Status: RESOLVED FIXED

(Testing :: General, defect, P2)

Description

[Dave Hunt [:davehunt] [he/him] ⌚BST]

As mentioned in bug 1036675, let's change the currently automatically set memory configuration jobs to use 512MB. Due to the limited devices, we shouldn't try to run against this configuration for every b2g-inbound build and Gaia commit. Would once per day be enough, or should we try to run against every b2g-inbound build (but allow builds to be skipped when devices are not available)?

Comment 1

[Mike Lee [:mlee]]

(In reply to Dave Hunt (:davehunt) from comment #0)
> ...
> Would once per day be enough, or should we try to run
> against every b2g-inbound build (but allow builds to be skipped when devices
> are not available)?

The latter. Skipping builds as necessary to free up devices is acceptable as long as we can have at least 1 run per day. Multiple runs is preferred so we can directly and quickly compare results from the same builds against the 319MB config when needed.

Comment 2

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Okay, I've set the 319MB jobs to follow every b2g-inbound build and Gaia commit (and to queue builds so we shouldn't miss any). I've also repurposed the other jobs to set 512MB and run only when there's a new build (and to not queue builds). We should see the flame-512MB results appearing in datazilla soon, though I strongly suspect we'll see bug 1046056 again here.

Comment 3

[Dave Hunt [:davehunt] [he/him] ⌚BST]

This is done.