-
Type: Task
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: Build, Testing Infrastructure
-
Fully Compatible
-
v3.4
-
TIG 2017-06-19
-
0
There have been many instances where the OOM killer has killed a mongod process when running on the Code Coverage builder. With the num_jobs_available expansion equal to the number of CPUs, we end up running 4 tests concurrently on the rhel62-small distro is a c3.xlarge (4 CPU, 7.5GiB memory). We should change to use the rhel62-large distro and set num_jobs_available to number of CPUs divided by 2 when running tests. The rhel62-large distro is a c3.4xlarge (16 CPU, 30GiB memory), so we'll end up running 8 tests concurrently, but have a larger ratio (2x) of available memory to number of concurrent tests.
Note: The special handling for running tasks on the rhel62-large distro can be removed since all tasks for the build variant will run on that distro anyway.