mirror of
https://github.com/hanzoai/tasks.git
synced 2026-10-04 02:04:03 +00:00
Vladyslav SimonenkoandGitHub
1916d05702
## What changed? Instrument gRPC dial calls with metrics monitoring connection establishment. This PR adds the following metrics: 1. ServiceDialLatency 2. ServiceDialSuccessCount 3. ServiceDialErrorCount ## Why? Currently the visibility into gRPC dial calls is limited, there are no metrics for how long establishing the connection took, or how many failures we encountered. Since `grpc.NewClient` is "lazy", in the sense that it doesn't establish connection right away, but rather on the first RPC, we need to instrument it via some hooks. This PR uses `grpc.WithContextDialer` to achieve this. gRPC library itself uses platform-specific keep alive hack to use OS defaults for TCP keepalive, but since we are on Go 1.23+ we can simply use newly introduced KeepAliveConfig. ## How did you test it? - [X] built - [X] run locally and tested manually - [X] covered by existing tests - [X] added new unit test(s) - [ ] added new functional test(s) ## Potential risks
87 lines
2.0 KiB
Go
87 lines
2.0 KiB
Go
package rpc
|
|
|
|
import (
|
|
"context"
|
|
"net/http/httptrace"
|
|
"testing"
|
|
"time"
|
|
|
|
"github.com/stretchr/testify/assert"
|
|
"github.com/stretchr/testify/require"
|
|
"go.temporal.io/server/common/log"
|
|
"go.temporal.io/server/common/metrics"
|
|
)
|
|
|
|
func TestDialTracer_DialSuccess(t *testing.T) {
|
|
t.Parallel()
|
|
|
|
addr := "127.0.0.1:1234"
|
|
dt := newDialTracer(addr, metrics.NoopMetricsHandler, log.NewNoopLogger())
|
|
|
|
ctx := context.Background()
|
|
ctx, ndt := dt.beginNetworkDial(ctx)
|
|
|
|
trace := httptrace.ContextClientTrace(ctx)
|
|
require.NotNil(t, trace)
|
|
|
|
trace.ConnectStart("tcp", addr)
|
|
rewind(&ndt.connectStart)
|
|
trace.ConnectDone("tcp", addr, nil)
|
|
|
|
dt.endNetworkDial(ndt, nil)
|
|
|
|
assert.Positive(t, ndt.connectDuration)
|
|
assert.Equal(t, addr, ndt.connectAddr)
|
|
assert.Nil(t, ndt.connectErr)
|
|
}
|
|
|
|
func TestDialTracer_DialError(t *testing.T) {
|
|
t.Parallel()
|
|
|
|
addr := "127.0.0.1:1234"
|
|
dt := newDialTracer(addr, metrics.NoopMetricsHandler, log.NewNoopLogger())
|
|
|
|
ctx := context.Background()
|
|
ctx, ndt := dt.beginNetworkDial(ctx)
|
|
|
|
trace := httptrace.ContextClientTrace(ctx)
|
|
require.NotNil(t, trace)
|
|
|
|
connectErr := assert.AnError
|
|
trace.ConnectStart("tcp", addr)
|
|
rewind(&ndt.connectStart)
|
|
trace.ConnectDone("tcp", addr, connectErr)
|
|
|
|
dt.endNetworkDial(ndt, nil)
|
|
|
|
assert.Positive(t, ndt.connectDuration)
|
|
assert.Equal(t, addr, ndt.connectAddr)
|
|
assert.Equal(t, connectErr, ndt.connectErr)
|
|
}
|
|
|
|
func TestDialTracer_NoConnectStart(t *testing.T) {
|
|
t.Parallel()
|
|
|
|
addr := "127.0.0.1:1234"
|
|
dt := newDialTracer(addr, metrics.NoopMetricsHandler, log.NewNoopLogger())
|
|
|
|
ctx := context.Background()
|
|
ctx, ndt := dt.beginNetworkDial(ctx)
|
|
|
|
trace := httptrace.ContextClientTrace(ctx)
|
|
require.NotNil(t, trace)
|
|
|
|
trace.ConnectDone("tcp", addr, nil)
|
|
|
|
dt.endNetworkDial(ndt, nil)
|
|
|
|
assert.Zero(t, ndt.connectDuration)
|
|
assert.Equal(t, addr, ndt.connectAddr)
|
|
assert.Nil(t, ndt.connectErr)
|
|
}
|
|
|
|
// rewind moves a timestamp back by 1 millisecond to ensure positive durations in tests
|
|
func rewind(ts *time.Time) {
|
|
*ts = ts.Add(-time.Millisecond)
|
|
}
|