[terraform] Increase Identity memory to 2048
ClosedPublic
Actions

Authored by varun on Jul 18 2024, 2:22 AM.

Details

Reviewers

will
ashoat
bartek

Commits

rCOMM2f2e98ec0869: [terraform] Increase Identity memory to 2048

Summary

Updated launch template to t3.xlarge (16GB RAM) - can consider also t3.large (8GB), I'll be experimenting with it on staging
Updated Identity task definition to use 2048 MB memory

Test Plan

Applied on staging, then prod.

Diff Detail

Repository

rCOMM Comm

Branch

barthap/tf-ecs

Lint

No Lint Coverage

Unit

No Test Coverage

Event Timeline

bartek created this revision.Jul 18 2024, 2:22 AM

bartek held this revision as a draft.

Herald added a subscriber: tomek. · View Herald TranscriptJul 18 2024, 2:22 AM

Harbormaster completed remote builds in B30457: Diff 42456.Jul 18 2024, 2:39 AM

Assuming this draft is ready for review since @bartek linked in on Linear

services/terraform/remote/aws_ecs.tf
74	What services use this config? Does identity use this config?
76	What does the change in `min_size` do?
77	In what contexts do we reach `max_size`? It seemed like we have 2 instances of identity when I was looking on the AWS console.

ashoat accepted this revision.Jul 18 2024, 6:16 AM

This revision is now accepted and ready to land.Jul 18 2024, 6:16 AM

In D12792#362215, @ashoat wrote:

Assuming this draft is ready for review since @bartek linked in on Linear

Yeah, I wasn't publishing it yet because of some experiments I was doing on staging (with rolling deployments when low on resources) but I think I'll stick with configuration from this diff, it's doing just fine.

services/terraform/remote/aws_ecs.tf
74	All EC2-based services, including Identity and Blob
76–77	I should've added comments here: These numbers determine the number of VMs able to run EC2-based services. Now it's set to be between 1-5 `t3.xlarge` VM instances (machines on which services run). It's not the number of particular service instances. One VM instance can host one or more service instances (tasks), depending on service resource consumption vs resources available on the VM. It was 0-3 before, but it felt outdated - we have 2 instances running constantly, and often a 3rd is needed when doing a rolling (zero-downtime) deploy. We never have 0 anymore - this value was left from times, where we had no services deployed yet (during initial setup) to avoid wasting resources.