Run task and wait for result in AWS ECS

ecs-run

#!/usr/bin/env bash
set -e

function usage() {
    set -e
    cat <<EOM
    ##### ecs-run #####
    Simple script for running tasks on Amazon Elastic Container Service
    One of the following is required:
    Required arguments:
        -d | --task-definition       Name of task definition to deploy
        -c | --cluster               Name of ECS cluster
        -n | --container-name        Name of Docker container

    Optional arguments:
        -m | --command
        --aws-instance-profile  Use the IAM role associated with this instance
        -D | --desired-count    The number of instantiations of the task to place.
        -t | --retries          Default is 12 for two hours. Script monitors ECS Service for new task definition to be running.
        -e | --tag-env-var      Get image tag name from environment variable. If provided this will override value specified in image name argument.
        -v | --verbose          Verbose output
        -r | --region           AWS Region
        -p | --profile          AWS Profile to use
    Requirements:
        aws:  AWS Command Line Interface
        jq:   Command-line JSON processor
    Examples:
      Simple deployment of a service (Using env vars for AWS settings):
        ecs-run -c production1 -d foo-taskdef -n foo-container -m "sleep,15"
      All options:
EOM

    exit 2
}
if [ $# == 0 ]; then usage; fi

function _echo {
    echo "${PREFIX}$@"
}

function command_to_array {
    local plain_array=""
    local IFS=' '
    for item in $@; do
        plain_array=$plain_array\"$item\"
    done
    echo "[$(echo $plain_array | sed 's/,/","/g')]"
}

# Check requirements
function require {
    command -v $1 > /dev/null 2>&1 || {
        _echo "Some of the required software is not installed:"
        _echo "    please install $1" >&2;
        exit 1;
    }
}

function logs_link {
    local service_name=$1
    local task_id=$(echo $2 | cut -d / -f 2)
    local logs_name=$(echo $service_name | cut -d \- -f 1)
    echo "https://eu-west-1.console.aws.amazon.com/cloudwatch/home?region=eu-west-1#logEventViewer:group=awslogs-${logs_name};stream=${service_name}/${service_name}/${task_id}"
}

# Check for AWS, AWS Command Line Interface
require aws
# Check for jq, Command-line JSON processor
require jq

# Setup default values for variables
CLUSTER=false
TASK_DEFINITION=false
CONTAINER_NAME=false
COMMAND=false
VERBOSE=false
TAGVAR=false
AWS_CLI=$(which aws)
if [ -n "$DRY_RUN" ]; then
    AWS_ECS="echo $AWS_CLI --output json ecs"
else
    AWS_ECS="$AWS_CLI --output json ecs"
fi


DESIRED=1

# Retry the command if the reason of failure is REOURCE:CPU or RESOURCE:MEMORY
MAX_RETRIES=5
RETRY_SLEEP_TIME=60

RETRIES_ACCEPTED_FAILURES=(
    RESOURCE:CPU
    RESOURCE:MEMORY
)

# Two hours waiting
MAX_WAITER_RETRIES=12

# Loop through arguments, two at a time for key and value
while [[ $# > 0 ]]
do
    key="$1"

    case $key in
        -p|--profile)
            AWS_PROFILE="$2"
            shift # past argument
            ;;
        --aws-instance-profile)
            AWS_IAM_ROLE=true
            ;;
        -c|--cluster)
            CLUSTER="$2"
            shift # past argument
            ;;
        -n|--container-name)
            CONTAINER_NAME="$2"
            shift # past argument
            ;;
        -d|--task-definition)
            TASK_DEFINITION="$2"
            shift
            ;;
        -m|--command)
            COMMAND="$(command_to_array $2)"
            shift
            ;;
        -r|--region)
            AWS_DEFAULT_REGION="$2"
            shift # past argument
            ;;
        -D|--desired-count)
            DESIRED="$2"
            shift
            ;;
        -e|--tag-env-var)
            TAGVAR="$2"
            shift
            ;;
        -v|--verbose)
            VERBOSE=true
            shift
            ;;
        -t|--retries)
            MAX_WAITER_RETRIES="$2"
            shift
            ;;
        *)
            _echo "ERROR: $1 is not a valid option"
            usage
            exit 2
        ;;
    esac
    shift # past argument or value
done

if [ $VERBOSE == true ]; then
    set -x
fi

if [ $TASK_DEFINITION == false ]; then
    _echo "TASK DEFINITON is required. You can pass the value using -d / --task-definiton for a task"
    exit 1
fi
if [ $CLUSTER == false ]; then
    _echo "CLUSTER is required. You can pass the value using -c or --cluster"
    exit 1
fi
if [ $CONTAINER_NAME == false ]; then
    _echo "CONTAINER_NAME is required. You can pass the value using -n or --container-name"
    exit 1
fi
if [ "$COMMAND" == false ]; then
    OVERRIDES="{\"containerOverrides\": [{\"name\": \"$CONTAINER_NAME\"}]}"
else
    OVERRIDES='{"containerOverrides": [{"name": "'"$CONTAINER_NAME"'", "command": '"$COMMAND"'}]}'
fi
if [ -z ${AWS_DEFAULT_REGION+x} ]; then
    unset AWS_DEFAULT_REGION
else
    AWS_ECS="$AWS_ECS --region $AWS_DEFAULT_REGION"
fi
if [ -z ${AWS_PROFILE+x} ]; then
    unset AWS_PROFILE
else
    AWS_ECS="$AWS_ECS --profile $AWS_PROFILE"
fi

run_task () {
    local run_result
    run_result=$($AWS_ECS run-task \
                          --cluster $CLUSTER \
                          --task-definition $TASK_DEFINITION \
                          --count $DESIRED \
                          --overrides "$OVERRIDES")
    local returned_value=$?
    echo $run_result
    return $returned_value
}

DO_RETRY=1
RETRIES=0


set +e
while [ "$RETRIES" -lt $MAX_RETRIES ] && [ $DO_RETRY -eq 1 ]; do
  DO_RETRY=0
  REASON_FAILURE=''
  RUN_TASK=$(run_task)
  RUN_TASK_EXIT_CODE=$?

  _echo $RUN_TASK

  FAILURES=$(echo $RUN_TASK | jq '.failures|length')
  if [ $FAILURES -eq 0 ]; then
    TASK_ARN=$(echo $RUN_TASK | jq '.tasks[0].taskArn' | sed -e 's/^"//' -e 's/"$//')
    WAITER_RETRY=1
    WAITER_RETRIES=0
    while [ $WAITER_RETRIES -lt $MAX_WAITER_RETRIES ] && [ $WAITER_RETRY -eq 1 ]; do
        WAITER_RETRY=0
        $AWS_ECS wait tasks-stopped --tasks "$TASK_ARN" --cluster $CLUSTER 2>/dev/null
        WAITER_EXIT_CODE=$?

        if [ $WAITER_EXIT_CODE -eq 0 ]; then
            DESCRIBE_TASKS=$($AWS_ECS describe-tasks --tasks "$TASK_ARN" --cluster $CLUSTER)
            EXIT_CODE=$(echo $DESCRIBE_TASKS | jq '.tasks[0].containers[0].exitCode')
            if [ $EXIT_CODE -eq 0 ]; then
              _echo "ECS task exited successfully"
              _echo $(logs_link $CONTAINER_NAME $TASK_ARN)
              exit 0
            else
              _echo "ECS task failed: $DESCRIBE_TASKS"
             _echo $(logs_link $CONTAINER_NAME $TASK_ARN)
              exit $EXIT_CODE
            fi

        elif [ $WAITER_EXIT_CODE -eq 255 ]; then
            ((WAITER_RETRIES++))
            WAITER_RETRY=1
            if [ $WAITER_RETRIES -eq $MAX_WAITER_RETRIES ]; then
                _echo "ECS Waiter max retries reached, $WAITER_RETRIES, exit"
                exit 255
            fi
            _echo "ECS Waiter because timeout,  waiter retry $WAITER_RETRIES (don't launch the task other time)"
        else
            _echo "ECS Waiter failed, status: $WAITER_EXIT_CODE"
            _echo $(logs_link $CONTAINER_NAME $TASK_ARN)
            exit $WAITER_EXIT_CODE
        fi
    done
  else
    REASON_FAILURE=$(echo $RUN_TASK | jq -r '.failures[0].reason')
    if [ -n "$REASON_FAILURE" ] && [[ "${RETRIES_ACCEPTED_FAILURES[@]}" =~ $REASON_FAILURE ]]; then
        DO_RETRY=1
        ((RETRIES++))
        if [ -n "$REASON_FAILURE" ] && [ $RETRIES -eq $MAX_RETRIES ]; then
            _echo "Max RETRIES reached REASON: $REASON_FAILURE  RETRIES: $RETRIES"
            _echo $(logs_link $CONTAINER_NAME $TASK_ARN)
            exit 253
        fi
        _echo "Retrying in ${RETRY_SLEEP_TIME}s, try number: $RETRIES because: $REASON_FAILURE"
        sleep $RETRY_SLEEP_TIME
    else
        _echo "ECS task failed: $REASON_FAILURE"
        _echo $(logs_link $CONTAINER_NAME $TASK_ARN)
        exit 1
    fi
  fi
done

Can I ask how the TAGVAR variable should be used? I am interested in how to override a docker image in run-task command, but I don't see the variable used anywhere.

@Victorcoder why do you retry on those errors (RESOURCE:CPU and RESOURCE:MEMORY)? Is your cluster on an AutoScaling Group?

You need to delete line 143 in order to get --verbose flag working correctly

This is great. I found it via SO. I updated this script to accommodate Fargate launch types: https://gist.github.com/sblack4/2edc4f64693792b57963d8c99d12e95e