| 1 |
# Exporting Metrics and Traces with OpenCensus, Zipkin, and Prometheus |
| 2 |
|
| 3 |
This tutorial provides a minimum example to verify that metrics and traces |
| 4 |
can be exported to OpenCensus from Go tools. |
| 5 |
|
| 6 |
## Setting up oragent |
| 7 |
|
| 8 |
1. Ensure you have [docker](https://www.docker.com/get-started) and [docker-compose](https://docs.docker.com/compose/install/). |
| 9 |
2. Clone [oragent](https://github.com/orijtech/oragent). |
| 10 |
3. In the oragent directory, start the services: |
| 11 |
```bash |
| 12 |
docker-compose up |
| 13 |
``` |
| 14 |
If everything goes well, you should see output resembling the following: |
| 15 |
``` |
| 16 |
Starting oragent_zipkin_1 ... done |
| 17 |
Starting oragent_oragent_1 ... done |
| 18 |
Starting oragent_prometheus_1 ... done |
| 19 |
... |
| 20 |
``` |
| 21 |
* You can check the status of the OpenCensus agent using zPages at http://localhost:55679/debug/tracez. |
| 22 |
* You can now access the Prometheus UI at http://localhost:9445. |
| 23 |
* You can now access the Zipkin UI at http://localhost:9444. |
| 24 |
4. To shut down oragent, hit Ctrl+C in the terminal. |
| 25 |
5. You can also start oragent in detached mode by running `docker-compose up -d`. To stop oragent while detached, run `docker-compose down`. |
| 26 |
|
| 27 |
## Exporting Metrics and Traces |
| 28 |
1. Clone the [tools](https://golang.org/x/tools) subrepository. |
| 29 |
1. Inside `internal`, create a file named `main.go` with the following contents: |
| 30 |
```go |
| 31 |
package main |
| 32 |
|
| 33 |
import ( |
| 34 |
"context" |
| 35 |
"fmt" |
| 36 |
"math/rand" |
| 37 |
"net/http" |
| 38 |
"time" |
| 39 |
|
| 40 |
"golang.org/x/tools/internal/event" |
| 41 |
"golang.org/x/tools/internal/event/export" |
| 42 |
"golang.org/x/tools/internal/event/export/metric" |
| 43 |
"golang.org/x/tools/internal/event/export/ocagent" |
| 44 |
) |
| 45 |
|
| 46 |
type testExporter struct { |
| 47 |
metrics metric.Exporter |
| 48 |
ocagent *ocagent.Exporter |
| 49 |
} |
| 50 |
|
| 51 |
func (e *testExporter) ProcessEvent(ctx context.Context, ev event.Event) (context.Context, event.Event) { |
| 52 |
ctx, ev = export.Tag(ctx, ev) |
| 53 |
ctx, ev = export.ContextSpan(ctx, ev) |
| 54 |
ctx, ev = e.metrics.ProcessEvent(ctx, ev) |
| 55 |
ctx, ev = e.ocagent.ProcessEvent(ctx, ev) |
| 56 |
return ctx, ev |
| 57 |
} |
| 58 |
|
| 59 |
func main() { |
| 60 |
exporter := &testExporter{} |
| 61 |
|
| 62 |
exporter.ocagent = ocagent.Connect(&ocagent.Config{ |
| 63 |
Start: time.Now(), |
| 64 |
Address: "http://127.0.0.1:55678", |
| 65 |
Service: "go-tools-test", |
| 66 |
Rate: 5 * time.Second, |
| 67 |
Client: &http.Client{}, |
| 68 |
}) |
| 69 |
event.SetExporter(exporter) |
| 70 |
|
| 71 |
ctx := context.TODO() |
| 72 |
mLatency := event.NewFloat64Key("latency", "the latency in milliseconds") |
| 73 |
distribution := metric.HistogramFloat64Data{ |
| 74 |
Info: &metric.HistogramFloat64{ |
| 75 |
Name: "latencyDistribution", |
| 76 |
Description: "the various latencies", |
| 77 |
Buckets: []float64{0, 10, 50, 100, 200, 400, 800, 1000, 1400, 2000, 5000, 10000, 15000}, |
| 78 |
}, |
| 79 |
} |
| 80 |
|
| 81 |
distribution.Info.Record(&exporter.metrics, mLatency) |
| 82 |
|
| 83 |
for { |
| 84 |
sleep := randomSleep() |
| 85 |
_, end := event.StartSpan(ctx, "main.randomSleep()") |
| 86 |
time.Sleep(time.Duration(sleep) * time.Millisecond) |
| 87 |
end() |
| 88 |
event.Record(ctx, mLatency.Of(float64(sleep))) |
| 89 |
|
| 90 |
fmt.Println("Latency: ", float64(sleep)) |
| 91 |
} |
| 92 |
} |
| 93 |
|
| 94 |
func randomSleep() int64 { |
| 95 |
var max int64 |
| 96 |
switch modulus := time.Now().Unix() % 5; modulus { |
| 97 |
case 0: |
| 98 |
max = 17001 |
| 99 |
case 1: |
| 100 |
max = 8007 |
| 101 |
case 2: |
| 102 |
max = 917 |
| 103 |
case 3: |
| 104 |
max = 87 |
| 105 |
case 4: |
| 106 |
max = 1173 |
| 107 |
} |
| 108 |
return rand.Int63n(max) |
| 109 |
} |
| 110 |
|
| 111 |
``` |
| 112 |
3. Run the new file from within the tools repository: |
| 113 |
```bash |
| 114 |
go run internal/main.go |
| 115 |
``` |
| 116 |
4. After about 5 seconds, OpenCensus should start receiving your new metrics, which you can see at http://localhost:8844/metrics. This page will look similar to the following: |
| 117 |
``` |
| 118 |
# HELP promdemo_latencyDistribution the various latencies |
| 119 |
# TYPE promdemo_latencyDistribution histogram |
| 120 |
promdemo_latencyDistribution_bucket{vendor="otc",le="0"} 0 |
| 121 |
promdemo_latencyDistribution_bucket{vendor="otc",le="10"} 2 |
| 122 |
promdemo_latencyDistribution_bucket{vendor="otc",le="50"} 9 |
| 123 |
promdemo_latencyDistribution_bucket{vendor="otc",le="100"} 22 |
| 124 |
promdemo_latencyDistribution_bucket{vendor="otc",le="200"} 35 |
| 125 |
promdemo_latencyDistribution_bucket{vendor="otc",le="400"} 49 |
| 126 |
promdemo_latencyDistribution_bucket{vendor="otc",le="800"} 63 |
| 127 |
promdemo_latencyDistribution_bucket{vendor="otc",le="1000"} 78 |
| 128 |
promdemo_latencyDistribution_bucket{vendor="otc",le="1400"} 93 |
| 129 |
promdemo_latencyDistribution_bucket{vendor="otc",le="2000"} 108 |
| 130 |
promdemo_latencyDistribution_bucket{vendor="otc",le="5000"} 123 |
| 131 |
promdemo_latencyDistribution_bucket{vendor="otc",le="10000"} 138 |
| 132 |
promdemo_latencyDistribution_bucket{vendor="otc",le="15000"} 153 |
| 133 |
promdemo_latencyDistribution_bucket{vendor="otc",le="+Inf"} 15 |
| 134 |
promdemo_latencyDistribution_sum{vendor="otc"} 1641 |
| 135 |
promdemo_latencyDistribution_count{vendor="otc"} 15 |
| 136 |
``` |
| 137 |
5. After a few more seconds, Prometheus should start displaying your new metrics. You can view the distribution at http://localhost:9445/graph?g0.range_input=5m&g0.stacked=1&g0.expr=rate(oragent_latencyDistribution_bucket%5B5m%5D)&g0.tab=0. |
| 138 |
|
| 139 |
6. Zipkin should also start displaying traces. You can view them at http://localhost:9444/zipkin/?limit=10&lookback=300000&serviceName=go-tools-test. |