deploy¶
The deploy section allows to set various settings around how the container should be deployed, and what compute resources are required to run the service.
For more details on the deploy, see docker documentation for deploy here
At the moment, all keys are not supported, mostly due to the way Fargate by nature is expecting settings to be.
resources¶
The resources is probably what interests most individuals, in setting up how much CPU and RAM should be setup for the service. I have tried to capture for various exceptions for the RAM settings, as you can find in ecs_composex.ecs.docker_tools.set_memory_to_mb
Once the container definitions are put together, the CPU and RAM requirements are put together. From there, it will automatically select the closest valid Fargate CPU/RAM combination and set the parameter for the Task definition.
Important
CPUs should be set between 0.25 and 4 to be valid for Fargate, otherwise you will have an error.
Warning
At the moment, I decided to hardcode these values in the CFN template. It is ugly, but pending bigger work to allow services merging, after which these will be put into a CFN parameter to allow you to change it on the fly.
replicas¶
This setting allows you to define how many tasks should be running for a given service. To make this work, I simply update the MicroserviceCount parameter default value, to keep things configurable.
Note
update_config will be use very soon to support replacement of services using a LB to possibly use CodeDeploy Blue/Green deployment.
labels¶
These labels aren’t used for much in native Docker compose as per the documentation. They are only used for the service, but not for the containers themselves. Which is great for us, as we can then leverage that structure to implement a merge of services.
In AWS ECS, a Task definition is a group of one or more containers which are going to be running as a one task. The most usual use-case for this, is with web applications, which need to have a reverse proxy (ie. nginx) in front of the actual application. But also, if you used the use_xray option, you realized that ECS ComposeX automatically adds the x-ray-daemon sidecar. Equally, when we implement AppMesh, we will also have another side-car container for this.
So, here is the tag that will allow you to merge your reverse proxy or waf (if you used a WAF in container) fronting your web application:
ecs.task.family¶
For example, you would have:
---
# base file for services with the x-keys for BDD
version: '3.8'
secrets:
abcd: {}
john:
x-secrets:
LinksTo:
- EcsExecutionRole
- EcsTaskRole
Name: SFTP/asl-cscs-files-dev
zyx:
x-secrets:
Name: secret/with/kmskey
Lookup:
Tags:
- costcentre: lambda
JsonKeys:
- VarName: ZYX_TEST
SecretKey: test
services:
app01:
env_file: ./use-cases/env-files/dummy.env
deploy:
labels:
ecs.task.family: bignicefamily
resources:
reservations:
cpus: '0.25'
memory: 1GB
environment:
LOGLEVEL: DEBUG
SHELLY: /bin/bash
TERMY: screen
image: nginx
links:
- app03:dateteller
ports:
- mode: awsvpc
protocol: tcp
published: 5000
target: 5000
secrets:
- zyx
x-logging:
RetentionInDays: 42
CreateLogGroup: False
x-network:
is_public: False
use_cloudmap: True
Ingress:
Myself: False
AwsSources:
- Type: PrefixList
Id: pl-6da54004
x-iam:
Policies:
- PolicyName: AllowPublishToCw
PolicyDocument:
Statement:
- Action:
- cloudwatch:PutMetricData
Effect: Allow
Resource:
- '*'
Sid: AllowPublishMetricsToCw
x-xray: false
x-scaling:
Range: "1-4"
app02:
env_file:
- ./use-cases/env-files/dummy.env
deploy:
labels:
ecs.task.family: youtoo
replicas: 2
resources:
reservations:
cpus: '0.1'
memory: 64000kB
environment:
LOGLEVEL: DEBUG
healthcheck:
interval: 30
retries: 3
test:
- CMD
- curl
- localhost:5000/ping
image: nginx
ports:
- mode: awsvpc
protocol: tcp
published: 5000
target: 5000
secrets:
- zyx
volumes:
- source: some-volume
target: /app/data
type: volume
x-iam:
PermissionsBoundary: arn:aws:iam::aws:policy/AdministratorAccess
ManagedPolicyArns:
- arn:aws:iam::aws:policy/AdministratorAccess
x-scaling:
Range: "1-5"
TargetScaling:
CpuTarget: 88
DisableScaleIn: true
x-xray: false
app03:
deploy:
resources:
reservations:
cpus: '0.25'
memory: 134217728b
environment:
LOGLEVEL: DEBUG
image: nginx
ports:
- mode: awsvpc
protocol: tcp
published: 5000
target: 5000
secrets:
- abcd
- zyx
- john
volumes:
- normal-vol
- shared-images:/app/images
- some-volume:/app/data:ro
x-network:
Ingress:
Myself: False
ExtSources:
- Ipv4: 0.0.0.0/0
Description: ANYWHERE
x-logging:
RetentionInDays: 30
x-scaling:
Range: 1-10
rproxy:
depends_on:
- app01
- app02
deploy:
labels:
ecs.task.family: bignicefamily,youtoo
replicas: 1
resources:
limits:
cpus: '0.25'
memory: 64M
reservations:
cpus: '0.1'
memory: 32M
image: nginx
ports:
- mode: awsvpc
protocol: tcp
published: 80
target: 80
x-iam:
ManagedPolicyArns:
- arn:aws:iam::aws:policy/ReadOnlyAccess
x-xray: true
x-network:
is_public: False
use_cloudmap: True
volumes:
shared-images: {}
some-volume: {}
normal-vol: {}
x-dns:
PrivateNamespace:
Name: lambda.internal
x-tags:
costcentre: lambda
Warning
The example above illustrates that you can either use, for deploy labels
a list of strings
a dictionary
ecs.depends.condition¶
This label allows to define what condition should this service be monitored under by ECS. Useful when container is set as a dependency to another.
Hint
Allowed values are : START, SUCCESS, COMPLETE, HEALTHY. By default, sets to START, and if you defined healthcheck, defaults to HEALTHY. See Dependency reference for more information