conf

Documentation

Overview

Package conf represents configuration of the clusters set.

Index

• Constants
• func ClustersHash(clusters *Clusters) (string, error)
• func Hash(cfg *Main, clusters *Clusters, rules *Rules, rewrites *Rewrites) (string, error)
• type Cluster
• type Clusters
  • func ReadClustersConfig(r io.Reader) (Clusters, error)
• type Host
• type Main
  • func MakeDefault() Main
  • func ReadMain(r io.Reader) (Main, error)
• type Rewrite
• type Rewrites
  • func ReadRewrites(r io.Reader) (Rewrites, error)
• type Rule
• type Rules
  • func ReadRules(r io.Reader) (Rules, error)

Constants

const BlackholeCluster string = "blackhole"

BlackholeCluster is used for testing. It cannot be specified in config. It eats the records, they are not sent anywhere.

const JumpCluster string = "jump"

JumpCluster distributes the incoming datapoints between hosts according to fnv1a hasing following by the jump one. Every record is sent to only one host.

const LB string = "lb"

LB cluster type distributes incoming datapoints using fnv modulo N, where N equal no. of available hosts As such, this cluster type keeps the list of available hosts and sends traffic only to these

const ToallCluster string = "toall"

ToallCluster broadcasts records to all hosts in the cluster.

Functions

func ClustersHash

func ClustersHash(clusters *Clusters) (string, error)

ClustersHash calculates hash of the clusters config to track its versions.

func Hash

func Hash(cfg *Main, clusters *Clusters, rules *Rules, rewrites *Rewrites) (string, error)

Hash calculates hash of all the configs to track config versions.

Types

type Cluster

type Cluster struct {
	Name  string `toml:"name"`
	Type  string `toml:"type"`
	Hosts []Host `toml:"hosts"`
}

Cluster is the config of a single cluster. Cluster type can be either *toall* or *jump*.

type Clusters

type Clusters struct {
	Cluster []Cluster
}

Clusters is the config for all the clusters in the system.

func ReadClustersConfig

func ReadClustersConfig(r io.Reader) (Clusters, error)

ReadClustersConfig reads clusters set from a reader.

type Host

type Host struct {
	// Name must be an FQDN or IP.
	Name string `toml:"name"`
	// Only makes sense when there is hashing.
	Index int `toml:"index"`
	// Optional, can be taken from default config.
	Port uint16 `toml:"port"`
	// Optional, TCP by default
	GRPC bool `toml:"grpc"`
	// Optional, number of TCP connections to open to the host.
	MTCP int `toml:"mtcp"`
}

Host is the config of a single machine to send metrics to. The index is set explicitly for easy modification of config and to make it more robust. Port is optional, set to default if absent.

type Main

type Main struct {
	// Clusters config path. Mandatory.
	ClustersConfig string
	// Rules config path. Mandatory.
	RulesConfig string
	// Rewrites config path. Optional.
	RewritesConfig string

	// Launch NT as a k8s daemonset forwarding metrics from pods.
	K8sMode bool
	// Use filter pods by label?
	K8sUseK8sServer bool
	// Pods to inject for TCP listening on selected pods. Metics sent to that port will be forwarded
	// to the running daemon-set.
	K8sInjectPortTCP uint16
	// The label to use in order to turn on forwarding from a pod.
	K8sSwitchLabelKey string
	// K8s label value. Default is "enabled".
	K8sSwitchLabelVal string
	// The period for updating the containers for metrics forwarding in k8s.
	K8sContainerUpdPeriodSec int
	// Range of jitter applied when querying k8s API while looking for pods.
	// Jitter added will be [0, this_number]
	K8sObserveJitterRangeSec int

	// The default target port on receiver hosts. Can be overridden in the clusters setup.
	TargetPort uint16

	// Setting this to ip:port will enable TCP listening on specified address.
	// Use empty string to disable TCP listening.
	// Use empty ip to listen on all addresses.
	ListenTCP string
	// Setting this to ip:port will enable UDP listening on specified address.
	// Use the empty string to disable UDP listening.
	// Use the empty IP to listen on all addresses.
	ListenUDP string
	// Setting this to ip:port will enable gRPC listening on specified address.
	// Use the empty string to disable gRPC listening.
	// Use the empty IP to listen on all addresses.
	ListenGRPC string

	// The size on main queue between listen and routing. Refer to docs for details.
	MainQueueSize uint64
	// Each host has it's own queue that contains records to be sent to it. This is the size of it.
	// Refer to docs for more insight.
	HostQueueSize uint64
	// The size of records batch sent into main queue.
	MainQueueBatchSize uint64
	// Period to flush the buffer when making batches for the main q.
	BatchFlushPerdiodSec uint32

	// Number of workers processing main queue and putting records into host queues.
	// If 0, set to be GOMAXPROCS / 2.
	// Default: 0.
	WorkerPoolSize int

	// Timeout for dropping an incoming connection if no data is sent in.
	IncomingConnIdleTimeoutSec uint32

	// Timeout for sending data to target host via TCP connection.
	// Has to be > TCPOutBufFlushPeriodSec to allow enough time for the buffered records to be sent.
	SendTimeoutSec uint32
	// Timeout for connecting to a target host. Does not influence re-connections with exponential backoffs.
	OutConnTimeoutSec uint32

	MaxHostReconnectPeriodMs   uint32
	HostReconnectPeriodDeltaMs uint32
	KeepAliveSec               uint32
	// Time to wait for the processing pipeline to terminate when quitting. After this timeout is passed,
	// forced termination is done. This helps when graceful shutdown is stuck or slow.
	TermTimeoutSec uint16
	// The size of the buffer for sending data to a target host via a TCP connections.
	// 0 value turns off buffering
	TCPOutBufSize int
	// The period over which the out TCP buffer for the connections sending to target hosts is flushed.
	// This helps if the traffic is low and records get stuck in the buffer.
	// The value of zero means no flushing.
	// 0 value turns off flushing
	TCPOutBufFlushPeriodSec uint32

	// The period to refresh the outgoing TCP connection.
	// This solves the following problem. If nothing is sent for a prolonged period of time to a client
	// it may drop the connection because of read i/o timeout. This will go unnoticed until the first
	// write attempt that will fail and data will be lost. To prevent this situation and data loss, connection
	// can be dropped and established from scratch.
	// 0 = no connection refresh.
	TCPOutConnectionRefreshPeriodSec uint32
	// Perform connection check to target right away?
	TCPInitialConnCheck bool

	// Enable jitter when connecting to targets?
	TargetConnectionJitter bool
	// Minimal amplitude when adding jitter to target connections. Has to be >0.
	TargetConnectionJitterMinAmplitudeMs uint32

	// gRPC target (client) params
	// gRPC HTTP2 keepalive ping period https://github.com/grpc/grpc-go/blob/master/Documentation/keepalive.md
	GRPCOutKeepAlivePeriodSec      uint32
	GRPCOutKeepAlivePingTimeoutSec uint32
	// Max connect backoff period https://github.com/grpc/grpc/blob/master/doc/connection-backoff.md
	GRPCOutBackoffMaxDelaySec uint32
	// Min time for the connection to complete https://github.com/grpc/grpc/blob/master/doc/connection-backoff.md
	GRPCOutMinConnectTimeoutSec uint32

	// gRPC listener (server) params
	GRPCListenMaxConnectionIdleSec     uint32
	GRPCListenMaxConnectionAgeSec      uint32
	GRPCListenMaxConnectionAgeGraceSec uint32
	GRPCListenTimeSec                  uint32
	GRPCListenTimeoutSec               uint32

	GRPCTracing bool

	// Turns on and off the normalization of records path. Described in the docs in detail.
	NormalizeRecords bool
	// Turns on logging for special kinds of records. For now it's recrods with fractional timestamps.
	LogSpecialRecords bool

	// -1 turns off pprof server
	PprofPort int
	PromPort  int
	// Switch to expose only small subset of essential metrics.
	// (Useful to reduce Prometheus load when running as a sidecar on many nodes in a large setup.)
	LessMetrics bool
	// Expose prometheus metrics with the total time for running each regex from config.
	// Can be used to understand what regexs from config are more 'expensive'
	RegexDurationMetric bool

	// Absolute path for a pidfile. Not written if left empty.
	PidFilePath string

	// Initial number of allowed log records during LogLimitWindowSec with same msg and level
	LogLimitInitial int
	// Log every LogLimitThereafter record during LogLimitWindowSec after initial set with same msg and level
	LogLimitThereafter int
	// Timeframe for log limiting
	LogLimitWindowSec int

	// Histogram parameters for the host queue size
	HostQueueLengthBucketFactor float64
	HostQueueLengthBuckets      int

	// Histogram parameters for the processing duration
	ProcessingDurationBucketFactor float64
	ProcessingDurationBuckets      int

	// Limits the number of records in the time window a container can send to maximum of
	// RateLimiterContainerRecordLimit records. Setting this to zero, disables container rate limiters.
	RateLimiterContainerRecordLimit int
	// Limits the number of records in the time window that generally can be sent to maximum of
	// RateLimiterGlobalRecordLimit records. Setting this to zero, disables the global rate limiter.
	RateLimiterGlobalRecordLimit int
	// Time window size of rate limiter.
	RateLimiterWindowSizeSec int
	// Record threshold for updating rate limiter in a reader. If set to zero or lower, rate limiter is
	// updated on each record
	RateLimiterPerReaderRecordThreshold int
	// Interval in which rate limiter is updated. If set equal or lower than zero, rate limiter updates only on record
	// threshold.
	RateLimiterIntervalMs int
	// Duration in which rate limiter hangs until trying again.
	RateLimiterRetryDurationMs int
}

Main is the main and generic nanotube config.

func MakeDefault

func MakeDefault() Main

MakeDefault creates configuration with default values.

func ReadMain

func ReadMain(r io.Reader) (Main, error)

ReadMain reads the main config

type Rewrite

type Rewrite struct {
	From string
	To   string
	Copy bool
}

Rewrite in configuration for a single rewrite rule.

type Rewrites

type Rewrites struct {
	Rewrite []Rewrite
}

Rewrites represents rewrites config.

func ReadRewrites

func ReadRewrites(r io.Reader) (Rewrites, error)

ReadRewrites reads rules from the reader. Errors when parsing fails.

type Rule

type Rule struct {
	Regexs   []string
	Prefixes []string
	Clusters []string
	Continue bool
}

Rule in configuration for a single rule.

type Rules

type Rules struct {
	Rule []Rule
}

Rules represents rules config.

func ReadRules

func ReadRules(r io.Reader) (Rules, error)

ReadRules reads rules from the reader. Errors when parsing fails.