oteltrpc

README

TRPC-Go metric Reporting

The following is the data reported by opentelemetry in trpc-go, which is convenient for users to query in Grafana and create panels using PromQL.

Metrics

Since some attributes are common to many metrics, we extract some common indicators here to avoid lengthy documentation. The following common attribute names will be used to replace these duplicate attrs.

• common_attrs:
  • app: ${app} in trpc.yaml
  • server: ${server}
  • namespace: ${namespace}
  • env_name: ${env_name}
  • container_name: ${container_name}
• rpc_relation:
  • system_name: hard code: "trpc"
  • caller_service: msg.CallerServiceName()
  • caller_method: msg.CallerMethod()
  • callee_service: msg.CalleeServiceName()
  • callee_method: msg.CalleeMethod()

callee metrics

server_started_total

• type: counter
• desc:
• attrs:
  • common_attrs
  • rpc_relation

server_handled_total

• type: counter
• desc: total number of requests received and processed by the server
• attrs:
  • common_attrs
  • rpc_relation
  • code: ret code
  • code_type: ret code type
    • success
    • exception
    • timeout
  • code_desc: ret code description

server_handled_seconds

• type: histogram
• desc: distribution of request processing time
• bucket: Supports custom buckets, default buckets are [0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 5] (counting the number of requests less than 5ms, 10ms, 25ms, 50ms, 100ms, 250ms, 0.5s, 1s, 5s, respectively)
• attrs:
  • common_attrs
  • rpc_relation

caller metrics

client_started_total

• type: counter
• desc: total number of requests sent
• attrs:
  • common_attrs
  • rpc_relation

client_handled_total

• type: counter
• desc: total number of requests sent and completed (may receive a response or timeout)
• attrs:
  • common_attrs
  • rpc_relation
  • code:
  • code_type:
    • success
    • exception
    • timeout
  • code_desc:

client_handled_seconds

• type: histogram
• desc: distribution of request processing time
• bucket: Supports custom buckets, default buckets are [0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 5] (counting the number of requests less than 5ms, 10ms, 25ms, 50ms, 100ms, 250ms, 0.5s, 1s, 5s, respectively)
• attrs:
  • common_attrs
  • rpc_relation

System

cpu_cores

• type: gauge
• desc: cpu cores(the correct quota of the container can be obtained within the container_
• frequency: only reported when server started
• attrs:
  • common_attrs

server_panic_total

• type: counter
• desc: panic counter
• attrs:
  • common_attrs
  • system_name: hard code "trpc"

user defined metrics

• Supports reporting of single-dimensional and multi-dimensional metric data for trpc-go.
• For the three data types of counter, gauge, and histogram, the metric name will be prefixed with trpc_counter_, trpc_gauge_, and trpc_histogram_, respectively.
• The uniqueness of the metric name for the same type of business indicator must be ensured, otherwise, it will trigger a panic.

Trace

Resource

all trace include following metric：

• tps.tenant.id: tenent id
• telemetry.sdk.language: hardcode "go"
• telemetry.sdk.name: hardcode "opentelemetry"
• service.name(optional): {app}_{server} in trpc_go.yaml

common attributes

common attributes for caller and callee:

• net:
  • peer:
    • ip: caller ip
    • port: caller port
  • host:
    • ip: callee ip
    • port: callee port
    • name: callee hostname
• trpc:
  • caller_service: msg.CallerServiceName
  • caller_method: msg.CallerMethod
  • callee_service: msg.CalleeServiceName
  • callee_method: msg.CalleeMethod
  • namespace: msg.Namespace (Development | Production)
• tps:
  • dyeing: dyeing key, msg.DyeingKey
  • force.sample(optional): flag for force sampling

callee

• common attributes
• badage
• traceContext
• spanKind: SPAN_KIND_SERVER
• spanName: msg.ServerRPCName

caller

• common attributes
• spanKind: SPAN_KIND_CLIENT
• spanName: msg.ClientRPCName

Documentation

Index

• Variables
• func Register()
• func RegisterTextMapSupplier(carrier func(md codec.MetaData, msg codec.Msg) propagation.TextMapCarrier)

Variables

var ClientFilter = filter.ClientChain{traces.ClientFilter(), prometheus.ClientFilter()}.Filter

var DefaultRecoveryHandler = func(ctx context.Context, panicErr interface{}) error {
	return fmt.Errorf("panic:%v", panicErr)
}

DefaultRecoveryHandler default recovery handler can be set by user

var DefaultSampler sdktrace.Sampler

DefaultSampler sampler could be set by user

var ParseConfig = func(configDec plugin.Decoder) (*config.Config, error) {
	cfg := &config.Config{}
	*cfg = config.DefaultConfig()
	if err := configDec.Decode(cfg); err != nil {
		return nil, err
	}
	return cfg, nil
}

ParseConfig can be set by the user to override the config

var ServerFilter = filter.ServerChain{traces.ServerFilter(), prometheus.ServerFilter(), logs.LogRecoveryFilter()}.Filter

Functions

func Register

func Register()

Register plugin and filter

func RegisterTextMapSupplier

func RegisterTextMapSupplier(carrier func(md codec.MetaData, msg codec.Msg) propagation.TextMapCarrier)

RegisterTextMapSupplier .