Add GPU instance support to the `up` command #743

efekarakus · 2019-03-08T19:06:54Z

Summary

Issue #, if available: 729

Description of changes: The ecs-cli up command now fetched the recommended Amazon ECS GPU optimized AMI metadata given a GPU instance type.

Testing

Ran make docker-test and the unit tests pass

Manual

Ran the command ./bin/local/ecs-cli up --capability-iam --instance-type p2.xlarge --cluster-config no-gpu-support --launch-type EC2 --force and can confirm that the EC2 instance uses the AMI ID amzn2-ami-ecs-gpu-hvm-2.0.20190301-x86_64-ebs (ami-0ebf2c738e66321e6)

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

ecs-cli/modules/cli/cluster/cluster_app.go

ecs-cli/modules/clients/aws/amimetadata/client.go

allisaurus

Can you please provide some test output that shows ecs-cli up running successfully with:

a GPU instance
an arm64 instance
a less specialized instance type (e.g., t2 family)
no instance type specified

ecs-cli/modules/cli/cluster/cluster_app.go

efekarakus · 2019-03-11T17:24:55Z

Manual Testing

New functionality

p2.xlarge

./bin/local/ecs-cli up --capability-iam --instance-type p2.xlarge --launch-type EC2 —force

Successful cluster creation, and verified in console that the instance has a GPU optimized image:

WARN[0000] You will not be able to SSH into your EC2 instances without a key pair. 
INFO[0000] Using GPU ecs-optimized AMI because instance type was p2.xlarge 
INFO[0001] Using recommended Amazon Linux 2 AMI with ECS Agent 1.26.0 and Docker version 18.06.1-ce 
INFO[0001] Created cluster                               cluster=default region=us-east-1
INFO[0001] Waiting for your CloudFormation stack resources to be deleted... 
INFO[0001] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0062] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0124] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0185] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0216] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
INFO[0277] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
INFO[0338] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
VPC created: vpc-06b66bb4219a9e0c0
Security Group created: sg-03e2af850152fce2b
Subnet created: subnet-0e14561045cde0ac9
Subnet created: subnet-00421aceefc4c52cb
Cluster creation succeeded.

Existing functionality

a1.medium

./bin/local/ecs-cli up --capability-iam --instance-type a1.medium --launch-type EC2 —force

Successful cluster creation, and verified in console that the instance has an ARM optimized image:

WARN[0000] You will not be able to SSH into your EC2 instances without a key pair. 
INFO[0000] Using Arm ecs-optimized AMI because instance type was a1.medium 
INFO[0001] Using recommended Amazon Linux 2 AMI with ECS Agent 1.26.0 and Docker version 18.06.1-ce 
INFO[0001] Created cluster                               cluster=default region=us-east-1
INFO[0001] Waiting for your CloudFormation stack resources to be deleted... 
INFO[0001] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0033] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
INFO[0094] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
INFO[0155] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
VPC created: vpc-0f6c784332f7657ad
Security Group created: sg-03933bd25b4d445dc
Subnet created: subnet-0c108f8f7f7117b9c
Subnet created: subnet-03c43207ecfe984d3
Cluster creation succeeded.

a1.medium in unsupported region

./bin/local/ecs-cli up --capability-iam --instance-type a1.medium --region us-west-1 --launch-type EC2 --force

Successfully fails:

WARN[0000] You will not be able to SSH into your EC2 instances without a key pair. 
INFO[0000] Using Arm ecs-optimized AMI because instance type was a1.medium 
FATA[0000] Error executing 'up': Could not find Recommended Amazon Linux 2 AMI /aws/service/ecs/optimized-ami/amazon-linux-2/arm64/recommended in us-west-1; the AMI may not be supported in this region: ParameterNotFound: 
	status code: 400, request id: 69ad3e1c-7643-4971-a276-46774ecb34d7

t2.micro

./bin/local/ecs-cli up --capability-iam --instance-type t2.micro --cluster-config no-gpu-support --launch-type EC2 —force

Successful creation of cluster, and verified in console that the instance is created properly:

WARN[0000] You will not be able to SSH into your EC2 instances without a key pair. 
INFO[0001] Using recommended Amazon Linux 2 AMI with ECS Agent 1.26.0 and Docker version 18.06.1-ce 
INFO[0001] Created cluster                               cluster=no-gpu-support region=us-east-1
INFO[0001] Waiting for your CloudFormation stack resources to be deleted... 
INFO[0001] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0062] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0123] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0186] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
INFO[0247] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
INFO[0308] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
VPC created: vpc-02ba225c17d0e856d
Security Group created: sg-05283b77b9aa150e8
Subnet created: subnet-054bab4f149655ca7
Subnet created: subnet-040ce52a313c56e10
Cluster creation succeeded.

No instance type specified

./bin/local/ecs-cli up --capability-iam --launch-type EC2 —force

Successfully defaults to t2.micro:

WARN[0000] You will not be able to SSH into your EC2 instances without a key pair. 
INFO[0000] Defaulting instance type to t2.micro         
INFO[0001] Using recommended Amazon Linux 2 AMI with ECS Agent 1.26.0 and Docker version 18.06.1-ce 
INFO[0001] Created cluster                               cluster=default region=us-east-1
INFO[0001] Waiting for your CloudFormation stack resources to be deleted... 
INFO[0001] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0063] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0123] Cloudformation stack status                   stackStatus=DELETE_IN_PROGRESS
INFO[0186] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
INFO[0247] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
INFO[0308] Cloudformation stack status                   stackStatus=CREATE_IN_PROGRESS
VPC created: vpc-0c765334b7fd1745d
Security Group created: sg-0ac3d886102f0976d
Subnet created: subnet-0c290f945aa9d318a
Subnet created: subnet-0f4209d27e4e9f79a
Cluster creation succeeded.

SoManyHs · 2019-03-12T22:57:47Z

ecs-cli/modules/cli/cluster/cluster_app.go

@@ -359,7 +347,6 @@ func createCluster(context *cli.Context, awsClients *AWSClients, commandConfig *
 		return err
 	}

-	logrus.Info("Waiting for your cluster resources to be created...")


Any reason we're getting rid of this log statement?

I have no idea why I removed that log statement, sorry! I put it back :)

Refer to aws#729

Bump version to 1.13.0

2d98b5e

efekarakus requested review from PettitWesley, SoManyHs and allisaurus March 8, 2019 19:06

PettitWesley reviewed Mar 8, 2019

View reviewed changes

ecs-cli/modules/cli/cluster/cluster_app.go Show resolved Hide resolved

PettitWesley reviewed Mar 8, 2019

View reviewed changes

ecs-cli/modules/clients/aws/amimetadata/client.go Show resolved Hide resolved

allisaurus reviewed Mar 9, 2019

View reviewed changes

ecs-cli/modules/cli/cluster/cluster_app.go Show resolved Hide resolved

PettitWesley and others added 5 commits March 11, 2019 16:00

Support container instance tagging via tags in user data

01dc62c

Bump version to 1.13.1

78c4de9

Upgrade aws-sdk-go to v1.18.0

2c9a257

Upgrade revision of mock and re-generate mocks

cea6754

Remove support for go1.8 for travis-ci

c8b53d0

SoManyHs approved these changes Mar 12, 2019

View reviewed changes

Add GPU instance support to the up command

5d6fabd

Refer to aws#729

efekarakus force-pushed the gpu-support branch from ebba242 to 5d6fabd Compare March 13, 2019 16:00

efekarakus merged commit 5d6fabd into aws:gpu Mar 13, 2019

efekarakus deleted the gpu-support branch March 13, 2019 21:01

efekarakus mentioned this pull request Mar 18, 2019

Merge gpu branch into dev #751

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add GPU instance support to the `up` command #743

Add GPU instance support to the `up` command #743

efekarakus commented Mar 8, 2019

allisaurus left a comment

efekarakus commented Mar 11, 2019 •

edited

Loading

SoManyHs Mar 12, 2019

efekarakus Mar 13, 2019

Add GPU instance support to the up command #743

Add GPU instance support to the up command #743

Conversation

efekarakus commented Mar 8, 2019

Summary

Testing

Manual

allisaurus left a comment

Choose a reason for hiding this comment

efekarakus commented Mar 11, 2019 • edited Loading

Manual Testing

New functionality

p2.xlarge

Existing functionality

a1.medium

a1.medium in unsupported region

t2.micro

No instance type specified

SoManyHs Mar 12, 2019

Choose a reason for hiding this comment

efekarakus Mar 13, 2019

Choose a reason for hiding this comment

Add GPU instance support to the `up` command #743

Add GPU instance support to the `up` command #743

efekarakus commented Mar 11, 2019 •

edited

Loading