docker_swarm_service_deployment.md

Docker Swarm Service Deployment

Deploy the services

Go the master node and run below to start 1 instance of a given container.

docker service create --replicas 1 --name alpine_app alpine ping docker.com

# --detach Exit immediately instead of waiting for the service to converge
docker service create --detach=false --name phpwebapp --publish published=8080,target=80 manojkmhub/phpapp

Verify the service.

docker service ls

docker service ps alpine_app

docker service ps alpine_app --format '{{.Node}}'

Inspect the service.

docker service inspect --pretty alpine_app

docker service inspect alpine_app

Check the nodes where services are running.

docker service ps alpine_app

Go the node where the service is running and verify.

docker container ls

docker ps

Scale the services

docker service scale alpine_app=5

Check the service details now. It is possible that services got started on master node too.

docker service ls

docker service ps alpine_app

You can scale down too.

docker service scale alpine_app=1

You can scale down to 0 as well. When you do this, service is not removed from swarm configuration. It is just that no service is running.

docker service scale alpine_app=0

docker service ls

You can scale up directlty from 0 to desired value.

docker service scale alpine_app=2

Delete the servics

docker service rm alpine_app

Update the service in rolling fashion

Services can be updated in many ways. See full list as below.

docker service update --help

Here, we will first create 4 containers of redis of some lower version. When we later upgrade the redis version, it will update the services one by one at interval defined by --update-delay.

By default, when once a service is updated, it will return RUNNING status, then next service will be updated after the given delay. If any update fails, it will pause there.

--update-delay
    seconds Ts
    minutes Tm
    hours Th

Example:
10s
10m30s -> 10 minute 30 seconds delay

Now, create the containers with lower version.

docker service create \
  --replicas 4 \
  --name redis_app \
  --update-delay 20s \
  redis:3.0.6

Verify the services.

docker service ps redis_app

docker service inspect --pretty redis_app

Now, update the image version.

docker service update --image redis:3.0.7 redis_app

Verify the previous and new specs of the service.

docker service inspect --format '{{.PreviousSpec.TaskTemplate.ContainerSpec.Image}}' redis_app

docker service inspect --format '{{.Spec.TaskTemplate.ContainerSpec.Image}}' redis_app

Another example of update:

docker service update --image redis:3.0.7 --update-delay 15s --update-order "start-first" redis_app

In next example, we will add an env var to the service.

docker service update --detach=false --env-add MY_NODE_ID="{{.Node.ID}}" --update-delay 20s redis_app

docker service ps redis_app

Verify the env var was updated.

docker service ps --format 'table {{.Name}}\t{{.Node}}\t{{.CurrentState}}' redis_app

# run this on the node where container has been started recently
docker container inspect --format '{{json .Config.Env}}' $(docker container ls -lq) | jq '.'

To restart a paused update, run docker service update . For example:

docker service update redis_app

To rollback the service update:

docker service update --detach=false --rollback redis_app

Rebalance the services

You can drain a nodes which will in effect rebalance the services among remaining active nodes.

Nodes can be in one of three availability states. ACTIVE, DRAIN, PAUSED. Default is ACTIVE. You can change the state from ACTIVE to DRAIN and vice versa.

When you change the node from ACTIVE to DRAIN state, any container running on that node as part of swarm set up will be stopped and launched on any ACTIVE nodes.

Check the node status and drain one of the nodes.

docker node ls

docker service ps redis_app

docker node update --availability drain ip-172-31-42-222

docker service ps redis_app --filter desired-state=running

docker service inspect redis_app

Make the node ACTIVE again. This will not rebalance the already running services across the nodes.

docker node update --availability active ip-172-31-42-222

But if you scale up, it will rebalance.

docker service ps redis_app --filter desired-state=running

docker service scale redis_app=5

docker service ps redis_app --filter desired-state=running

You can force the rebalance but it will stop and start the containers as needed.

docker service update --force redis_app