strategy

README

---

page_title: Docker Swarm strategies page_description: Swarm strategies page_keywords: docker, swarm, clustering, strategies

Strategies

The Docker Swarm scheduler features multiple strategies for ranking nodes. The strategy you choose determines how Swarm computes ranking. When you run a new container, Swarm chooses to place it on the node with the highest computed ranking for your chosen strategy.

To choose a ranking strategy, pass the --strategy flag and a strategy value to the swarm manage command. Swarm currently supports these values:

• spread
• binpack
• random

The spread and binpack strategies compute rank according to a node's available CPU, its RAM, and the number of containers it is running. The random strategy uses no computation. It selects a node at random and is primarily intended for debugging.

Your goal in choosing a strategy is to best optimize your swarm according to your company's needs.

Under the spread strategy, Swarm optimizes for the node with the fewest running containers. The binpack strategy causes Swarm to optimize for the node which is most packed. The random strategy, like it sounds, chooses nodes at random regardless of their available CPU or RAM.

Using the spread strategy results in containers spread thinly over many machines. The advantage of this strategy is that if a node goes down you only lose a few containers.

The binpack strategy avoids fragmentation because it leaves room for bigger containers on unused machines. The strategic advantage of binpack is that you use fewer machines as Swarm tries to pack as many containers as it can on a node.

If you do not specify a --strategy Swarm uses spread by default.

Spread strategy example

In this example, your swarm is using the spread strategy which optimizes for nodes that have the fewest containers. In this swarm, both node-1 and node-2 have 2G of RAM, 2 CPUs, and neither node is running a container. Under this strategy node-1 and node-2 have the same ranking.

When you run a new container, the system chooses node-1 at random from the swarm of two equally ranked nodes:

$ docker run -d -P -m 1G --name db mysql
f8b693db9cd6

$ docker ps
CONTAINER ID        IMAGE               COMMAND             CREATED                  STATUS              PORTS                           NODE        NAMES
f8b693db9cd6        mysql:latest        "mysqld"            Less than a second ago   running             192.168.0.42:49178->3306/tcp    node-1      db

Now, we start another container and ask for 1G of RAM again.

$ docker run -d -P -m 1G --name frontend nginx
963841b138d8

$ docker ps
CONTAINER ID        IMAGE               COMMAND             CREATED                  STATUS              PORTS                           NODE        NAMES
963841b138d8        nginx:latest        "nginx"             Less than a second ago   running             192.168.0.42:49177->80/tcp      node-2      frontend
f8b693db9cd6        mysql:latest        "mysqld"            Up About a minute        running             192.168.0.42:49178->3306/tcp    node-1      db

The container frontend was started on node-2 because it was the node with the fewest running containers. If two nodes have the same amount of available RAM and CPUs, the spread strategy prefers the node with the fewest containers running.

BinPack strategy example

In this example, let's says that both node-1 and node-2 have 2G of RAM and neither is running a container. Again, the nodes are equal. When you run a new container, the system chooses node-1 at random from the swarm:

$ docker run -d -P -m 1G --name db mysql
f8b693db9cd6

$ docker ps
CONTAINER ID        IMAGE               COMMAND             CREATED                  STATUS              PORTS                           NODE        NAMES
f8b693db9cd6        mysql:latest        "mysqld"            Less than a second ago   running             192.168.0.42:49178->3306/tcp    node-1      db

Now, you start another container, asking for 1G of RAM again.

$ docker run -d -P -m 1G --name frontend nginx
963841b138d8

$ docker ps
CONTAINER ID        IMAGE               COMMAND             CREATED                  STATUS              PORTS                           NODE        NAMES
963841b138d8        nginx:latest        "nginx"             Less than a second ago   running             192.168.0.42:49177->80/tcp      node-1      frontend
f8b693db9cd6        mysql:latest        "mysqld"            Up About a minute        running             192.168.0.42:49178->3306/tcp    node-1      db

The system starts the new frontend container on node-1 because it was the node with the most running containers. This allows us to start a container requiring 2G of RAM on node-2.

If two nodes have the same amount of available RAM and CPUs, the binpack strategy prefers the node with the most running containers.

Docker Swarm documentation index

• User guide
• Discovery options
• Scheduler filters
• Swarm API

Documentation

Index

• Variables
• func List() []string
• type BinpackPlacementStrategy
  • func (p *BinpackPlacementStrategy) Initialize() error
  • func (p *BinpackPlacementStrategy) Name() string
  • func (p *BinpackPlacementStrategy) RankAndSort(config *cluster.ContainerConfig, nodes []*node.Node) ([]*node.Node, error)
• type PlacementStrategy
  • func New(name string) (PlacementStrategy, error)
• type RandomPlacementStrategy
  • func (p *RandomPlacementStrategy) Initialize() error
  • func (p *RandomPlacementStrategy) Name() string
  • func (p *RandomPlacementStrategy) RankAndSort(config *cluster.ContainerConfig, nodes []*node.Node) ([]*node.Node, error)
• type SpreadPlacementStrategy
  • func (p *SpreadPlacementStrategy) Initialize() error
  • func (p *SpreadPlacementStrategy) Name() string
  • func (p *SpreadPlacementStrategy) RankAndSort(config *cluster.ContainerConfig, nodes []*node.Node) ([]*node.Node, error)

Variables

var (

	// ErrNotSupported is the error returned when a strategy name does not match
	// any supported placement strategy.
	ErrNotSupported = errors.New("strategy not supported")
	// ErrNoResourcesAvailable is the error returned when there are no resources
	// available to schedule a container. This can occur if there are no nodes in
	// the cluster or if no node contains sufficient resources for the container.
	ErrNoResourcesAvailable = errors.New("no resources available to schedule container")
)

Functions

func List

func List() []string

List returns the names of all the available strategies.

Types

type BinpackPlacementStrategy

type BinpackPlacementStrategy struct {
}

BinpackPlacementStrategy places a container onto the most packed node in the cluster.

func (*BinpackPlacementStrategy) Initialize

func (p *BinpackPlacementStrategy) Initialize() error

Initialize a BinpackPlacementStrategy.

func (*BinpackPlacementStrategy) Name

func (p *BinpackPlacementStrategy) Name() string

Name returns the name of the strategy.

func (*BinpackPlacementStrategy) RankAndSort

func (p *BinpackPlacementStrategy) RankAndSort(config *cluster.ContainerConfig, nodes []*node.Node) ([]*node.Node, error)

RankAndSort sorts nodes based on the binpack strategy applied to the container config.

type PlacementStrategy

type PlacementStrategy interface {
	// Name of the strategy
	Name() string
	// Initialize performs any initial configuration required by the strategy and returns
	// an error if one is encountered.
	// If no initial configuration is needed, this may be a no-op and return a nil error.
	Initialize() error
	// RankAndSort applies the strategy to a list of nodes and ranks them based
	// on the best fit given the container configuration.  It returns a sorted
	// list of nodes (based on their ranks) or an error if there is no
	// available node on which to schedule the container.
	RankAndSort(config *cluster.ContainerConfig, nodes []*node.Node) ([]*node.Node, error)
}

PlacementStrategy is the interface for a container placement strategy.

func New

func New(name string) (PlacementStrategy, error)

New creates a new PlacementStrategy for the given strategy name.

type RandomPlacementStrategy

type RandomPlacementStrategy struct {
	// contains filtered or unexported fields
}

RandomPlacementStrategy randomly places the container into the cluster.

func (*RandomPlacementStrategy) Initialize

func (p *RandomPlacementStrategy) Initialize() error

Initialize a RandomPlacementStrategy.

func (*RandomPlacementStrategy) Name

func (p *RandomPlacementStrategy) Name() string

Name returns the name of the strategy.

func (*RandomPlacementStrategy) RankAndSort

func (p *RandomPlacementStrategy) RankAndSort(config *cluster.ContainerConfig, nodes []*node.Node) ([]*node.Node, error)

RankAndSort randomly sorts the list of nodes.

type SpreadPlacementStrategy

type SpreadPlacementStrategy struct {
}

SpreadPlacementStrategy places a container on the node with the fewest running containers.

func (*SpreadPlacementStrategy) Initialize

func (p *SpreadPlacementStrategy) Initialize() error

Initialize a SpreadPlacementStrategy.

func (*SpreadPlacementStrategy) Name

func (p *SpreadPlacementStrategy) Name() string

Name returns the name of the strategy.

func (*SpreadPlacementStrategy) RankAndSort

func (p *SpreadPlacementStrategy) RankAndSort(config *cluster.ContainerConfig, nodes []*node.Node) ([]*node.Node, error)

RankAndSort sorts nodes based on the spread strategy applied to the container config.