balanced

Documentation

Overview

Package balanced implements an allocator that can sort allocations based on multiple metrics, where metrics may be an arbitrary way to partition a set of peers.

For example, allocating by ["tag:region", "disk"] the resulting peer candidate order will balanced between regions and ordered by the value of the weight of the disk metric.

Index

• Variables
• type Allocator
  • func New(cfg *Config) (*Allocator, error)
  • func (a *Allocator) Allocate(ctx context.Context, c api.Cid, current, candidates, priority api.MetricsSet) ([]peer.ID, error)
  • func (a *Allocator) Metrics() []string
  • func (a *Allocator) SetClient(c *rpc.Client)
  • func (a *Allocator) Shutdown(ctx context.Context) error
• type Config
  • func (cfg *Config) ApplyEnvVars() error
  • func (cfg *Config) ConfigKey() string
  • func (cfg *Config) Default() error
  • func (cfg *Config) LoadJSON(raw []byte) error
  • func (cfg *Config) ToDisplayJSON() ([]byte, error)
  • func (cfg *Config) ToJSON() ([]byte, error)
  • func (cfg *Config) Validate() error

Variables

var (
	DefaultAllocateBy = []string{"tag:group", "freespace"}
)

These are the default values for a Config.

Types

type Allocator

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

Allocator is an allocator that partitions metrics and orders the final list of allocation by selecting for each partition.

func New

func New(cfg *Config) (*Allocator, error)

New returns an initialized Allocator.

func (*Allocator) Allocate

func (a *Allocator) Allocate(
	ctx context.Context,
	c api.Cid,
	current, candidates, priority api.MetricsSet,
) ([]peer.ID, error)

Allocate produces a sorted list of cluster peer IDs based on different metrics provided for those peer IDs. It works as follows:

• First, it buckets each peer metrics based on the AllocateBy list. The metric name must match the bucket name, otherwise they are put at the end.
• Second, based on the AllocateBy order, it orders the first bucket and groups peers by ordered value.
• Third, it selects metrics on the second bucket for the most prioritary peers of the first bucket and orders their metrics. Then for the peers in second position etc.
• It repeats the process until there is no more buckets to sort.
• Finally, it returns the first peer of the first
• Third, based on the AllocateBy order, it select the first metric

func (*Allocator) Metrics

func (a *Allocator) Metrics() []string

Metrics returns the names of the metrics that have been registered with this allocator.

func (*Allocator) SetClient

func (a *Allocator) SetClient(c *rpc.Client)

SetClient provides us with an rpc.Client which allows contacting other components in the cluster.

func (*Allocator) Shutdown

func (a *Allocator) Shutdown(ctx context.Context) error

Shutdown is called on cluster shutdown. We just invalidate any metrics from this point.

type Config

type Config struct {
	config.Saver

	AllocateBy []string
}

Config allows to initialize the Allocator.

func (*Config) ApplyEnvVars

func (cfg *Config) ApplyEnvVars() error

ApplyEnvVars fills in any Config fields found as environment variables.

func (*Config) ConfigKey

func (cfg *Config) ConfigKey() string

ConfigKey returns a human-friendly identifier for this Config's type.

func (*Config) Default

func (cfg *Config) Default() error

Default initializes this Config with sensible values.

func (*Config) LoadJSON

func (cfg *Config) LoadJSON(raw []byte) error

LoadJSON parses a raw JSON byte-slice as generated by ToJSON().

func (*Config) ToDisplayJSON

func (cfg *Config) ToDisplayJSON() ([]byte, error)

ToDisplayJSON returns JSON config as a string.

func (*Config) ToJSON

func (cfg *Config) ToJSON() ([]byte, error)

ToJSON generates a human-friendly JSON representation of this Config.

func (*Config) Validate

func (cfg *Config) Validate() error

Validate checks that the fields of this configuration have sensible values.