Upgrade workers from m3 to c5d

Description

Gather some data and see if runtime is improved. Newrelic should help here. The c5d.large instances have newer, faster CPUs and directly attached NVME SSDs, but half the RAM of m3.large instances. Cost drops from $0.133/hour to $0.096/hour (and should finish faster to reduce the cost further, assuming the RAM isn't a bottleneck).

c5d instance type announcement: https://aws.amazon.com/about-aws/whats-new/2018/05/introducing-amazon-ec2-c5d-instances/
EC2 instance type spec and cost comparison: https://ec2instances.info/

Steps to Reproduce

None

Current Behavior

None

Expected Behavior

None

Reason for Variance

None

Release Notes

None

User Impact Summary

None

Activity

Show:
Jeremy Bowman
June 7, 2019, 2:55 PM

We might also want to upgrade github-status-worker from t2.micro to t3.micro; it should yield a modest performance gain, although the cost savings are negligible. I don't see a good upgrade path for coverage-worker away from m3.medium; the updated tiers no longer have a "medium" instance type, and all the cheaper options use EBS instead of a direct SSD.

Mike Youngstrom
June 3, 2019, 6:47 PM

Also look into the ECS containers. We had seen better performance gains by increasing containers rather than increasing processes per container, which is why we currently only run 1 xdist process on each. Maybe a newer instance type will change.

Done

Assignee

Mike Youngstrom

Reporter

Mike Youngstrom

Labels

None

Reach

None

Impact

None

Platform Area

None

Customer

None

Partner Manager

None

URL

None

Contributor Name

None

Groups with Read-Only Access

None

Actual Points

None

Category of Work

None

Platform Map Area (Levels 1 & 2)

None

Platform Map Area (Levels 3 & 4)

None

Story Points

1

Sprint

None

Priority

Unset