@cap-js/telemetry
is a CDS plugin providing observability features, including automatic OpenTelemetry instrumentation.
Documentation can be found at cap.cloud.sap and opentelemetry.io.
- About this project
- Requirements
- Setup
- Telemetry Signals
- Predefined Kinds
- Detailed Configuration Options
- Support, Feedback, Contributing
- Code of Conduct
- Licensing
See Getting Started on how to jumpstart your development and grow as you go with SAP Cloud Application Programming Model.
Simply add @cap-js/telemetry
to your dependencies via npm add @cap-js/telemetry
and you will find telemetry output written to the console like so:
[odata] - GET /odata/v4/processor/Incidents
[telemetry] - elapsed times:
0.00 → 2.85 = 2.85 ms GET /odata/v4/processor/Incidents
0.47 → 1.24 = 0.76 ms ProcessorService - READ ProcessorService.Incidents
0.78 → 1.17 = 0.38 ms db - READ ProcessorService.Incidents
0.97 → 1.06 = 0.09 ms @cap-js/sqlite - prepare SELECT json_object('ID',ID,'createdAt',createdAt,'creat…
1.10 → 1.13 = 0.03 ms @cap-js/sqlite - stmt.all SELECT json_object('ID',ID,'createdAt',createdAt,'crea…
1.27 → 1.88 = 0.61 ms ProcessorService - READ ProcessorService.Incidents.drafts
1.54 → 1.86 = 0.32 ms db - READ ProcessorService.Incidents.drafts
1.74 → 1.78 = 0.04 ms @cap-js/sqlite - prepare SELECT json_object('ID',ID,'DraftAdministrativeData_Dra…
1.81 → 1.85 = 0.04 ms @cap-js/sqlite - stmt.all SELECT json_object('ID',ID,'DraftAdministrativeData_Dr…
See Predefined Kinds for additional dependencies you need to bring yourself as well as some additional setup steps you need to perform when exporting to Dynatrace, SAP Cloud Logging, Jaeger, etc.
The plugin can be disabled by setting environment variable NO_TELEMETRY
to something truthy.
Database tracing is limited to @cap-js/cds-dbs-based databases, such as @cap-js/sqlite and @cap-js/hana.
There are three categories of telemetry data, also referred to as signals.
The following briefly describes, how each is addressed in @cap-js/telemetry
.
For more information on signals in general, please refer to https://opentelemetry.io/docs/concepts/signals.
Traces allow you to analyze how a request, message, task, etc. is being processed throughout your distributed system.
For this, @cap-js/telemetry
wraps all essential functions of cds.Service
and its derivates.
For @cap-js databases (e.g., @cap-js/sqlite
), this includes prepare()
and subsequent stmt.run()
and the likes.
An example trace printed to the console can be found in telemetry-to-console
.
In environments where Dynatrace OneAgent is installed (e.g., SAP BTP CF), no OpenTelemetry exporter is needed to transport the traces to Dynatrace.
@cap-js/telemetry
recognizes this and ignores any exporter config if the predefined kind telemetry-to-dynatrace
is used.
Metrics are "measurements captured at runtime", which help you understand your app's health and performance.
Out of the box, @cap-js/telemetry
offers metrics regarding the app's database pool, namely the pool info statistics of generic-pool
.
Example db pool metrics printed to the console:
[telemetry] - db.pool:
size | available | pending
1/1 | 1/1 | 0
Additionally, @cap-js/telemetry
instantiates and starts @opentelemetry/host-metrics
if it is found in the app's dependencies.
Currently, there is no public config option to influence which metrics @opentelemetry/host-metrics
collects.
However, it is possible to instruct the meter provider during initialization, which metrics shall be ignored.
By default, this is done for all system.*
metrics collected by @opentelemetry/host-metrics
.
This can be disabled via environment variable HOST_METRICS_RETAIN_SYSTEM=true
.
As these so-called views must be passed into the constructor, the above only applies in case @cap-js/telemetry
initializes the meter provider.
To avoid spamming the console, only process.*
metrics are printed by default, regardless of whether the system.*
metrics are ignored or not.
Printing the system.*
metrics (if not ignored) in the built-in console exporter can be enabled via environment variable HOST_METRICS_LOG_SYSTEM=true
.
Example host metrics printed to the console:
[telemetry] - host metrics:
Process Cpu time in seconds: { user: 1691.832, system: 218.223 }
Process Cpu usage time 0-1: { user: 82.07801878654074, system: 10.586932682237526 }
Process Memory usage in bytes: 141049856
Finally, custom metrics can be added as shown in the following example (tenant-aware request counting):
// server.js
const cds = require('@sap/cds')
let counter
cds.middlewares.add((req, _, next) => {
counter.add(1, { 'sap.tenancy.tenant_id': req.tenant })
next()
})
cds.on('listening', () => {
const { metrics } = require('@opentelemetry/api')
const meter = metrics.getMeter('@capire/incidents:req.counter')
counter = meter.createCounter('req.counter')
})
module.exports = cds.server
Exporting logs via OpenTelemetry is not yet supported.
There are three predefined kinds as follows:
Prints traces and metrics to the console as previously depicted (traces in Setup and metrics in Telemetry Signals - Metrics).
No additional dependencies are needed. This is the default kind in both development and production.
Exports traces and metrics to Dynatrace. Hence, a Dynatrace instance is required and the app must be bound to that Dynatrace instance.
Use via cds.requires.telemetry.kind = 'to-dynatrace'
.
Required additional dependencies:
@opentelemetry/exporter-trace-otlp-proto
(optional, see Leveraging Dynatrace OneAgent)@opentelemetry/exporter-metrics-otlp-proto
The necessary scopes for exporting traces (openTelemetryTrace.ingest
) and metrics (metrics.ingest
) are not part of the standard apitoken
and must be requested.
This can be done via parameterizing the binding to a "managed service instance" (i.e., not a user-provided service instance) as follows.
Excerpt from example mta.yaml:
requires:
- name: my-dynatrace-instance
parameters:
config:
tokens:
- name: ingest_apitoken #> default lookup name, configurable via cds.requires.telemetry.token_name
scopes:
- openTelemetryTrace.ingest
- metrics.ingest
In the user-provided service case, you'll need to generate a token in Dynatrace with the necessary scopes, add it to the credentials of the user-provided service, and configure cds.requires.telemetry.token_name
if the token's key in the credentials object is not ingest_apitoken
.
In Dynatrace itself, you need to ensure that the following two features are enabled:
- OpenTelemetry Node.js Instrumentation agent support:
- From the Dynatrace menu, go to Settings > Preferences > OneAgent features.
- Find and turn on OpenTelemetry Node.js Instrumentation agent support.
- W3C Trace Context:
- From the Dynatrace menu, go to Settings > Server-side service monitoring > Deep monitoring > Distributed tracing.
- Turn on Send W3C Trace Context HTTP headers.
If Dynatrace OneAgent is present, for example on SAP BTP CF, it will collect and transport the traces created by @cap-js/telemetry
automatically.
(Your app still needs to be bound to a Dynatrace instance, of course. However, @dynatrace/oneagent-sdk
is not required.)
Hence, additionally dependency @opentelemetry/exporter-trace-otlp-proto
and scope openTelemetryTrace.ingest
are not required.
This is actually the perferred operating model for telemetry-to-dynatrace
as it provides a better experience than exporting via OpenTelemetry.
If dependency @opentelemetry/exporter-trace-otlp-proto
is present anyway, @cap-js/telemetry
will export the traces via OpenTelemetry as well.
Exports traces and metrics to SAP Cloud Logging. Hence, a SAP Cloud Logging instance is required and the app must be bound to that SAP Cloud Logging instance.
Use via cds.requires.telemetry.kind = 'to-cloud-logging'
.
Required additional dependencies:
@grpc/grpc-js
@opentelemetry/exporter-trace-otlp-grpc
@opentelemetry/exporter-metrics-otlp-grpc
In order to receive OpenTelemetry credentials in the binding to the SAP Cloud Logging instance, you need to include the following configuration while creating the SAP Cloud Logging instance (or by updating an existing instance):
{
"ingest_otlp": {
"enabled": true
}
}
Exports traces to Jaeger.
Use via cds.requires.telemetry.kind = 'to-jaeger'
.
Required additional dependencies (As Jaeger does not support metrics, only a trace exporter is needed.):
@opentelemetry/exporter-trace-otlp-proto
Provide custom credentials like so:
{
"cds": {
"requires": {
"telemetry": {
"kind": "to-jaeger",
"tracing": {
"exporter": {
"config": { //> this object is passed into constructor as is
// add credentials here as decribed in
// https://www.npmjs.com/package/@opentelemetry/exporter-trace-otlp-proto
}
}
}
}
}
}
}
Run Jaeger locally via docker:
- Run
docker run -d --name jaeger -e COLLECTOR_ZIPKIN_HOST_PORT=:9411 -e COLLECTOR_OTLP_ENABLED=true -p 6831:6831/udp -p 6832:6832/udp -p 5778:5778 -p 16686:16686 -p 4317:4317 -p 4318:4318 -p 14250:14250 -p 14268:14268 -p 14269:14269 -p 9411:9411 jaegertracing/all-in-one:latest
- With this, no custom credentials are needed
- Open
localhost:16686
to see the traces
Exports traces and metrics to an OTLP/gRPC or OTLP/HTTP endpoint based on environment variables.
Use via cds.requires.telemetry.kind = 'to-otlp'
.
Required additional dependencies (* = grpc|proto|http
):
@grpc/grpc-js
(in case of OTLP/gRPC)@opentelemetry/exporter-trace-otlp-*
@opentelemetry/exporter-metrics-otlp-*
Please note that @cap-js/telemetry
does not validate the configuration via environment variables!
In general, you can influence the configuration of a used module via the respective config
node in cds.env.requires.telemetry
.
For example, it is possible to specify the temporalityPreference
setting of the respective metrics exporter like so:
{
"cds": {
"requires": {
"telemetry": {
"metrics": {
"exporter": {
"config": { //> this object is passed into constructor as is
"temporalityPreference": "DELTA"
}
}
}
}
}
}
}
Configure via cds.requires.telemetry.instrumentations = { <name>: { module, class, config? } }
Default:
{
"http": {
"module": "@opentelemetry/instrumentation-http",
"class": "HttpInstrumentation",
"config": {
"ignoreIncomingPaths": [
"/health"
]
}
}
}
Configure via cds.requires.telemetry.tracing.sampler = { kind, root?, ratio? }
Default:
{
"kind": "ParentBasedSampler",
"root": "AlwaysOnSampler"
}
Configure via cds.requires.telemetry.tracing.propagators = [<name> | { module, class, config? }]
Default:
["W3CTraceContextPropagator"]
Configure via:
cds.requires.telemetry.tracing.exporter = { module, class, config? }
cds.requires.telemetry.metrics.exporter = { module, class, config? }
Default:
{
{
"kind": "telemetry-to-console",
"tracing": {
"module": "@cap-js/telemetry",
"class": "ConsoleSpanExporter"
},
"metrics": {
"module": "@cap-js/telemetry",
"class": "ConsoleMetricExporter"
}
},
{
"kind": "telemetry-to-dynatrace",
"tracing": {
"exporter": {
"module": "@opentelemetry/exporter-trace-otlp-proto",
"class": "OTLPTraceExporter"
}
},
"metrics": {
"exporter": {
"module": "@opentelemetry/exporter-metrics-otlp-proto",
"class": "OTLPMetricExporter"
}
}
},
{
"kind": "telemetry-to-cloud-logging",
"tracing": {
"exporter": {
"module": "@opentelemetry/exporter-trace-otlp-grpc",
"class": "OTLPTraceExporter"
}
},
"metrics": {
"exporter": {
"module": "@opentelemetry/exporter-metrics-otlp-grpc",
"class": "OTLPMetricExporter"
}
}
},
{
"kind": "telemetry-to-jaeger",
"tracing": {
"exporter": {
"module": "@opentelemetry/exporter-trace-otlp-proto",
"class": "OTLPTraceExporter"
}
}
}
}
- For JSON output to the console, use:
{ "tracing": { "exporter": { "module": "@opentelemetry/sdk-trace-base", "class": "ConsoleSpanExporter" } }, "metrics": { "exporter": { "module": "@opentelemetry/sdk-metrics", "class": "ConsoleMetricExporter" } } }
- For HTTP, use:
{ "tracing": { "exporter": { "module": "@opentelemetry/exporter-trace-otlp-http", "class": "OTLPTraceExporter" } }, "metrics": { "exporter": { "module": "@opentelemetry/exporter-metrics-otlp-http", "class": "OTLPMetricExporter" } } }
By default, the start time of a span is taken from Date.now()
and, hence, has only millisecond resolution.
Via cds.requires.telemetry.tracing.hrtime = true
, you can instruct the plugin to specify the start and end times of spans, which it does with nanosecond resolution.
This may result in minor drifts, especially for spans created by other instrumentations such as @opentelemetry/instrumentation-http
.
Hence, the hrtime
mode is on by default in development but not in production.
NO_TELEMETRY
: Disables the pluginNO_LOCATE
: Disables function location in tracingSAP_PASSPORT
: Enables propagating W3C trace context to SAP HANA (experimental!)OTEL_LOG_LEVEL
: If not specified, the log level of cds loggertelemetry
is usedOTEL_SERVICE_NAME
: If not specified, the name is determined from package.json (defaulting to "CAP Application")OTEL_SERVICE_VERSION
: If not specified, the version is determined from package.json (defaulting to "1.0.0")
For the complete list of environment variables supported by OpenTelemetry, see Environment Variable Specification.
Please note that process.env.VCAP_APPLICATION
and process.env.CF_INSTANCE_GUID
, if present, are used to determine some Attributes.
This project is open to feature requests/suggestions, bug reports etc. via GitHub issues. Contribution and feedback are encouraged and always welcome. For more information about how to contribute, the project structure, as well as additional contribution information, see our Contribution Guidelines.
We as members, contributors, and leaders pledge to make participation in our community a harassment-free experience for everyone. By participating in this project, you agree to abide by its Code of Conduct at all times.
Copyright 2023 SAP SE or an SAP affiliate company and contributors. Please see our LICENSE for copyright and license information. Detailed information including third-party components and their licensing/copyright information is available via the REUSE tool.