monitor

package
v1.0.1-0...-a8ad18b Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Sep 21, 2016 License: MIT Imports: 15 Imported by: 0

README

System Monitoring

This functionality should be considered experimental and is subject to change.

System Monitoring means all statistical and diagnostic information made availabe to the user of InfluxDB system, about the system itself. Its purpose is to assist with troubleshooting and performance analysis of the database itself.

Statistics vs. Diagnostics

A distinction is made between statistics and diagnostics for the purposes of monitoring. Generally a statistical quality is something that is being counted, and for which it makes sense to store persistently for historical analysis. Diagnostic information is not necessarily numerical, and may not make sense to store.

An example of statistical information would be the number of points received over UDP, or the number of queries executed. Examples of diagnostic information would be a list of current Graphite TCP connections, the version of InfluxDB, or the uptime of the process.

System Statistics

SHOW STATS [FOR <module>] displays statisics about subsystems within the running influxd process. Statistics include points received, points indexed, bytes written to disk, TCP connections handled etc. These statistics are all zero when the InfluxDB process starts. If module is specified, it must be single-quoted. For example SHOW STATS FOR 'httpd'.

All statistics are written, by default, by each node to a "monitor" database within the InfluxDB system, allowing analysis of aggregated statistical data using the standard InfluxQL language. This allows users to track the performance of their system. Importantly, this allows cluster-level statistics to be viewed, since by querying the monitor database, statistics from all nodes may be queried. This can be a very powerful approach for troubleshooting your InfluxDB system and understanding its behaviour.

System Diagnostics

SHOW DIAGNOSTICS [FOR <module>] displays various diagnostic information about the influxd process. This information is not stored persistently within the InfluxDB system. If module is specified, it must be single-quoted. For example SHOW STATS FOR 'build'.

Standard expvar support

All statistical information is available at HTTP API endpoint /debug/vars, in expvar format, allowing external systems to monitor an InfluxDB node. By default, the full path to this endpoint is http://localhost:8086/debug/vars.

Configuration

The monitor module allows the following configuration:

  • Whether to write statistical and diagnostic information to an InfluxDB system. This is enabled by default.
  • The name of the database to where this information should be written. Defaults to _internal. The information is written to the default retention policy for the given database.
  • The name of the retention policy, along with full configuration control of the retention policy, if the default retention policy is not suitable.
  • The rate at which this information should be written. The default rate is once every 10 seconds.

Design and Implementation

A new module named monitor supports all basic statistics and diagnostic functionality. This includes:

  • Allowing other modules to register statistics and diagnostics information, allowing it to be accessed on demand by the monitor module.
  • Serving the statistics and diagnostic information to the user, in response to commands such as SHOW DIAGNOSTICS.
  • Expose standard Go runtime information such as garbage collection statistics.
  • Make all collected expvar data via HTTP, for collection by 3rd-party tools.
  • Writing the statistical information to the "monitor" database, for query purposes.

Registering statistics and diagnostics

To export statistical information with the monitor system, a service should implement the monitor.Reporter interface. Services added to the Server will be automatically added to the list of statistics returned. Any service that is not added to the Services slice will need to modify the Server's Statistics(map[string]string) method to aggregate the call to the service's Statistics(map[string]string) method so they are combined into a single response. The Statistics(map[string]string) method should return a statistics slice with the passed in tags included. The statistics should be kept inside of an internal structure and should be accessed in a thread-safe way. It is common to create a struct for holding the statistics and using sync/atomic instead of locking. If using sync/atomic, be sure to align the values in the struct so it works properly on i386.

To register diagnostic information, monitor.RegisterDiagnosticsClient is called, passing a influxdb.monitor.DiagsClient object to monitor. Implementing the influxdb.monitor.DiagsClient interface requires that your component have function returning diagnostic information in specific form, so that it can be displayed by the monitor system.

Statistical information is reset to its initial state when a server is restarted.

Documentation

Index

Constants

View Source
const (
	// DefaultStoreEnabled is whether the system writes gathered information in
	// an InfluxDB system for historical analysis.
	DefaultStoreEnabled = true

	// DefaultStoreDatabase is the name of the database where gathered information is written
	DefaultStoreDatabase = "_internal"

	// DefaultStoreInterval is the period between storing gathered information.
	DefaultStoreInterval = 10 * time.Second
)
View Source
const (
	MonitorRetentionPolicy         = "monitor"
	MonitorRetentionPolicyDuration = 7 * 24 * time.Hour
	MonitorRetentionPolicyReplicaN = 1
)

Policy constants.

Variables

This section is empty.

Functions

func DiagnosticsFromMap

func DiagnosticsFromMap(m map[string]interface{}) *diagnostics.Diagnostics

DiagnosticsFromMap returns a Diagnostics from a map.

Types

type Config

type Config struct {
	StoreEnabled  bool          `toml:"store-enabled"`
	StoreDatabase string        `toml:"store-database"`
	StoreInterval toml.Duration `toml:"store-interval"`
}

Config represents the configuration for the monitor service.

func NewConfig

func NewConfig() Config

NewConfig returns an instance of Config with defaults.

func (Config) Validate

func (c Config) Validate() error

Validate validates that the configuration is acceptable.

type Monitor

type Monitor struct {
	// Build information for diagnostics.
	Version   string
	Commit    string
	Branch    string
	BuildTime string

	MetaClient interface {
		CreateDatabaseWithRetentionPolicy(name string, spec *meta.RetentionPolicySpec) (*meta.DatabaseInfo, error)
		Database(name string) *meta.DatabaseInfo
	}

	// Writer for pushing stats back into the database.
	PointsWriter PointsWriter

	Logger *log.Logger
	// contains filtered or unexported fields
}

Monitor represents an instance of the monitor system.

func New

func New(r Reporter, c Config) *Monitor

New returns a new instance of the monitor system.

func (*Monitor) Close

func (m *Monitor) Close() error

Close closes the monitor system.

func (*Monitor) DeregisterDiagnosticsClient

func (m *Monitor) DeregisterDiagnosticsClient(name string)

DeregisterDiagnosticsClient deregisters a diagnostics client by name.

func (*Monitor) Diagnostics

func (m *Monitor) Diagnostics() (map[string]*diagnostics.Diagnostics, error)

Diagnostics fetches diagnostic information for each registered diagnostic client. It skips any clients that return an error when retrieving their diagnostics.

func (*Monitor) Open

func (m *Monitor) Open() error

Open opens the monitoring system, using the given clusterID, node ID, and hostname for identification purpose.

func (*Monitor) RegisterDiagnosticsClient

func (m *Monitor) RegisterDiagnosticsClient(name string, client diagnostics.Client)

RegisterDiagnosticsClient registers a diagnostics client with the given name and tags.

func (*Monitor) SetGlobalTag

func (m *Monitor) SetGlobalTag(key string, value interface{})

SetGlobalTag can be used to set tags that will appear on all points written by the Monitor.

func (*Monitor) SetLogOutput

func (m *Monitor) SetLogOutput(w io.Writer)

SetLogOutput sets the writer to which all logs are written. It must not be called after Open is called.

func (*Monitor) SetPointsWriter

func (m *Monitor) SetPointsWriter(pw PointsWriter) error

SetPointsWriter can be used to set a writer for the monitoring points.

func (*Monitor) Statistics

func (m *Monitor) Statistics(tags map[string]string) ([]*Statistic, error)

Statistics returns the combined statistics for all expvar data. The given tags are added to each of the returned statistics.

type PointsWriter

type PointsWriter interface {
	WritePoints(database, retentionPolicy string, points models.Points) error
}

PointsWriter is a simplified interface for writing the points the monitor gathers

type RemoteWriterConfig

type RemoteWriterConfig struct {
	RemoteAddr string
	NodeID     string
	Username   string
	Password   string
	ClusterID  uint64
}

RemoteWriterConfig represents the configuration of a remote writer

type Reporter

type Reporter interface {
	Statistics(tags map[string]string) []models.Statistic
}

type Statistic

type Statistic struct {
	models.Statistic
}

Statistic represents the information returned by a single monitor client.

func (*Statistic) ValueNames

func (s *Statistic) ValueNames() []string

valueNames returns a sorted list of the value names, if any.

type Statistics

type Statistics []*Statistic

func (Statistics) Len

func (a Statistics) Len() int

func (Statistics) Less

func (a Statistics) Less(i, j int) bool

func (Statistics) Swap

func (a Statistics) Swap(i, j int)

Directories

Path Synopsis

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL