Closed Bug 985650 Opened 8 years ago Closed 6 years ago

Allow certain jobs to run on beefier EC2 instances even if more pricey

Categories

(Release Engineering :: General, defect)

x86_64
Linux
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: armenzg, Unassigned)

References

Details

Android 2.3 on EC2 reftests would benefit with m3.large as per bug 980519.
B2g Reftests on EC2 run 15% to 60% slower than on the minis.

Even if we found solution to make them run faster, it will be good to have this flexibility.
To be clear, my argument in bug 980519 is that all Android 2.3 tests would benefit from m3.large, but particularly mochitests. mochitests on m3.large run in 1/3 the time they do currently on Ash. Also many of the mochitests which are currently disabled for Android 2.3 are skipped because of intermittent test timeouts, and I hope some of these mochitests could be run reliably on m3.large.

reftests on Android 2.3 are very slow currently. Running on m3.large, they run faster, but still not nearly fast enough to run in a reasonable number of chunks. We need to find other ways of improving reftest speed on Android 2.3 or run the emulator on real hardward (perhaps ix slaves).
I will enable the b2g reftests on our current m1.medium instances and run them side-by-side in the next 2 days.

Shutting the minis off would depend on 1) this bug or 2) we can't wait any longer on scl1.

The sooner we disable those minis the more we can save in the scl1 move out.
Note, the way to keep costs down is to use m1.medium ondemand nodes and bid on fast nodes on spot.
Blocks: 864866
No longer blocks: 818968
No longer blocks: 864866
Depends on: 864866
TC FTW!
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → WONTFIX
Component: General Automation → General
You need to log in before you can comment on or make changes to this bug.