Overlay network driver
overlay network driver creates a distributed network among multiple
Docker daemon hosts. This network sits on top of (overlays) the host-specific
networks, allowing containers connected to it (including swarm service
containers) to communicate securely when encryption is enabled. Docker
transparently handles routing of each packet to and from the correct Docker
daemon host and the correct destination container.
When you initialize a swarm or join a Docker host to an existing swarm, two new networks are created on that Docker host:
- an overlay network called
ingress, which handles the control and data traffic related to swarm services. When you create a swarm service and do not connect it to a user-defined overlay network, it connects to the
ingressnetwork by default.
- a bridge network called
docker_gwbridge, which connects the individual Docker daemon to the other daemons participating in the swarm.
You can create user-defined
overlay networks using
docker network create,
in the same way that you can create user-defined
bridge networks. Services
or containers can be connected to more than one network at a time. Services or
containers can only communicate across networks they are each connected to.
Although you can connect both swarm services and standalone containers to an overlay network, the default behaviors and configuration concerns are different. For that reason, the rest of this topic is divided into operations that apply to all overlay networks, those that apply to swarm service networks, and those that apply to overlay networks used by standalone containers.
Firewall rules for Docker daemons using overlay networks
You need the following ports open to traffic to and from each Docker host participating on an overlay network:
- TCP port 2377 for cluster management communications
- TCP and UDP port 7946 for communication among nodes
- UDP port 4789 for overlay network traffic
Before you can create an overlay network, you need to either initialize your Docker daemon as a swarm manager using
docker swarm initor join it to an existing swarm using
docker swarm join. Either of these creates the default
ingressoverlay network which is used by swarm services by default. You need to do this even if you never plan to use swarm services. Afterward, you can create additional user-defined overlay networks.
To create an overlay network for use with swarm services, use a command like the following:
$ docker network create -d overlay my-overlay
To create an overlay network which can be used by swarm services or
standalone containers to communicate with other standalone containers running on
other Docker daemons, add the
$ docker network create -d overlay --attachable my-attachable-overlay
You can specify the IP address range, subnet, gateway, and other options. See
docker network create --help for details.
All swarm service management traffic is encrypted by default, using the AES algorithmopen_in_new in GCM mode. Manager nodes in the swarm rotate the key used to encrypt gossip data every 12 hours.
To encrypt application data as well, add
--opt encrypted when creating the
overlay network. This enables IPSEC encryption at the level of the vxlan. This
encryption imposes a non-negligible performance penalty, so you should test this
option before using it in production.
When you enable overlay encryption, Docker creates IPSEC tunnels between all the nodes where tasks are scheduled for services attached to the overlay network. These tunnels also use the AES algorithm in GCM mode and manager nodes automatically rotate the keys every 12 hours.
Do not attach Windows nodes to encrypted overlay networks.
Overlay network encryption is not supported on Windows. If a Windows node attempts to connect to an encrypted overlay network, no error is detected but the node cannot communicate.
You can use the overlay network feature with both
--opt encrypted --attachable
and attach unmanaged containers to that network:
$ docker network create --opt encrypted --driver overlay --attachable my-attachable-multi-host-network
Most users never need to configure the
ingress network, but Docker allows you
to do so. This can be useful if the automatically-chosen subnet conflicts with
one that already exists on your network, or you need to customize other low-level
network settings such as the MTU.
ingress network involves removing and recreating it. This is
usually done before you create any services in the swarm. If you have existing
services which publish ports, those services need to be removed before you can
During the time that no
ingress network exists, existing services which do not
publish ports continue to function but are not load-balanced. This affects
services which publish ports, such as a WordPress service which publishes port
docker network inspect ingress, and remove any services whose containers are connected to it. These are services that publish ports, such as a WordPress service which publishes port 80. If all such services are not stopped, the next step fails.
Remove the existing
$ docker network rm ingress WARNING! Before removing the routing-mesh network, make sure all the nodes in your swarm run the same docker engine version. Otherwise, removal may not be effective and functionality of newly created ingress networks will be impaired. Are you sure you want to continue? [y/N]
Create a new overlay network using the
--ingressflag, along with the custom options you want to set. This example sets the MTU to 1200, sets the subnet to
10.11.0.0/16, and sets the gateway to
$ docker network create \ --driver overlay \ --ingress \ --subnet=10.11.0.0/16 \ --gateway=10.11.0.2 \ --opt com.docker.network.driver.mtu=1200 \ my-ingress
Note: You can name your
ingressnetwork something other than
ingress, but you can only have one. An attempt to create a second one fails.
Restart the services that you stopped in the first step.
docker_gwbridge is a virtual bridge that connects the overlay networks
ingress network) to an individual Docker daemon's physical
network. Docker creates it automatically when you initialize a swarm or join a
Docker host to a swarm, but it is not a Docker device. It exists in the kernel
of the Docker host. If you need to customize its settings, you must do so before
joining the Docker host to the swarm, or after temporarily removing the host
from the swarm.
Delete the existing
$ sudo ip link set docker_gwbridge down $ sudo ip link del dev docker_gwbridge
Start Docker. Do not join or initialize the swarm.
Create or re-create the
docker_gwbridgebridge manually with your custom settings, using the
docker network createcommand. This example uses the subnet
10.11.0.0/16. For a full list of customizable options, see Bridge driver options.
$ docker network create \ --subnet 10.11.0.0/16 \ --opt com.docker.network.bridge.name=docker_gwbridge \ --opt com.docker.network.bridge.enable_icc=false \ --opt com.docker.network.bridge.enable_ip_masquerade=true \ docker_gwbridge
Initialize or join the swarm. Since the bridge already exists, Docker does not create it with automatic settings.
Swarm services connected to the same overlay network effectively expose all
ports to each other. For a port to be accessible outside of the service, that
port must be published using the
--publish flag on
docker service create or
docker service update. Both the legacy colon-separated syntax and
the newer comma-separated value syntax are supported. The longer syntax is
preferred because it is somewhat self-documenting.
|-p 8080:80 or|
|Map TCP port 80 on the service to port 8080 on the routing mesh.|
|-p 8080:80/udp or|
|Map UDP port 80 on the service to port 8080 on the routing mesh.|
|-p 8080:80/tcp -p 8080:80/udp or|
-p published=8080,target=80,protocol=tcp -p published=8080,target=80,protocol=udp
|Map TCP port 80 on the service to TCP port 8080 on the routing mesh, and map UDP port 80 on the service to UDP port 8080 on the routing mesh.|
By default, swarm services which publish ports do so using the routing mesh.
When you connect to a published port on any swarm node (whether it is running a
given service or not), you are redirected to a worker which is running that
service, transparently. Effectively, Docker acts as a load balancer for your
swarm services. Services using the routing mesh are running in virtual IP (VIP)
mode. Even a service running on each node (by means of the
flag) uses the routing mesh. When using the routing mesh, there is no guarantee
about which Docker node services client requests.
To bypass the routing mesh, you can start a service using DNS Round Robin
(DNSRR) mode, by setting the
--endpoint-mode flag to
dnsrr. You must run
your own load balancer in front of the service. A DNS query for the service name
on the Docker host returns a list of IP addresses for the nodes running the
service. Configure your load balancer to consume this list and balance the
traffic across the nodes.
By default, control traffic relating to swarm management and traffic to and from
your applications runs over the same network, though the swarm control traffic
is encrypted. You can configure Docker to use separate network interfaces for
handling the two different types of traffic. When you initialize or join the
--datapath-addr separately. You must do
this for each node joining the swarm.
ingress network is created without the
--attachable flag, which means
that only swarm services can use it, and not standalone containers. You can
connect standalone containers to user-defined overlay networks which are created
--attachable flag. This gives standalone containers running on
different Docker daemons the ability to communicate without the need to set up
routing on the individual Docker daemon hosts.
|Map TCP port 80 in the container to port |
|Map UDP port 80 in the container to port |
|Map SCTP port 80 in the container to port |
|Map TCP port 80 in the container to TCP port |
For most situations, you should connect to the service name, which is load-balanced and handled by all containers ("tasks") backing the service. To get a list of all tasks backing the service, do a DNS lookup for
Due to limitations set by the Linux kernel, overlay networks become unstable and inter-container communications may break when 1000 containers are co-located on the same host.
For more information about this limitation, see moby/moby#44973open_in_new.