durability

Documentation

Overview

Package durability helps us to keep segments in healthy state with reporting risks.

Index

• type HealthStat
  • func (h *HealthStat) Merge(stat *HealthStat)
  • func (h *HealthStat) Min() int
  • func (h *HealthStat) Unused() bool
  • func (h *HealthStat) Update(num int)
• type NodeClassifier
• type ObserverFork
  • func (c *ObserverFork) Process(ctx context.Context, segments []rangedloop.Segment) (err error)
• type Report
  • func NewDurability(db overlay.DB, metabaseDB *metabase.DB, classifiers []NodeClassifier, ...) *Report
  • func (c *Report) Finish(ctx context.Context) error
  • func (c *Report) Fork(ctx context.Context) (rangedloop.Partial, error)
  • func (c *Report) Join(ctx context.Context, partial rangedloop.Partial) (err error)
  • func (c *Report) Start(ctx context.Context, startTime time.Time) error
  • func (c *Report) TestChangeReporter(r func(name string, stat *HealthStat))
• type ReportConfig

Types

type HealthStat

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

HealthStat collects the availability conditions for one class (for example: nodes with the same owner).

func (*HealthStat) Merge

func (h *HealthStat) Merge(stat *HealthStat)

Merge can merge two stat to one, without losing information.

func (*HealthStat) Min

func (h *HealthStat) Min() int

Min returns the minimal number.

func (*HealthStat) Unused

func (h *HealthStat) Unused() bool

Unused returns true when stat is uninitialized (-1) and was not updated with any number.

func (*HealthStat) Update

func (h *HealthStat) Update(num int)

Update updates the stat with one measurement: number of pieces which are available even without the nodes of the selected class.

type NodeClassifier

type NodeClassifier func(node *nodeselection.SelectedNode) string

NodeClassifier identifies a risk class (for example an owner, or country) of the SelectedNode.

type ObserverFork

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

ObserverFork is the durability calculator for each segment range.

func (*ObserverFork) Process

func (c *ObserverFork) Process(ctx context.Context, segments []rangedloop.Segment) (err error)

Process implements rangedloop.Partial.

type Report

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

Report is a calculator (rangloop.Observer) which checks the availability of pieces without certain nodes. It can answer the following question:

1. loosing a given group of nodes (all nodes of one country or all nodes of one owner)...
2. what will be the lowest humber of healhty pieces, checking all the segments.

Example: we have one segment where 80 pieces are stored, but 42 of them are in Germany.

in this case this reporter will return 38 for the class "country:DE" (assuming all the other segments are more lucky).

func NewDurability

func NewDurability(db overlay.DB, metabaseDB *metabase.DB, classifiers []NodeClassifier, reportThreshold int, asOfSystemInterval time.Duration) *Report

NewDurability creates the new instance.

func (*Report) Finish

func (c *Report) Finish(ctx context.Context) error

Finish implements rangedloop.Observer.

func (*Report) Fork

func (c *Report) Fork(ctx context.Context) (rangedloop.Partial, error)

Fork implements rangedloop.Observer.

func (*Report) Join

func (c *Report) Join(ctx context.Context, partial rangedloop.Partial) (err error)

Join implements rangedloop.Observer.

func (*Report) Start

func (c *Report) Start(ctx context.Context, startTime time.Time) error

Start implements rangedloop.Observer.

func (*Report) TestChangeReporter

func (c *Report) TestChangeReporter(r func(name string, stat *HealthStat))

TestChangeReporter modifies the reporter for unit tests.

type ReportConfig

type ReportConfig struct {
	Enabled bool `help:"whether to enable durability report (rangedloop observer)" default:"true"`
}

ReportConfig configures durability report.