Add support for InfluxDB

[steverao]

Resolves #1572

[laurit]

One thing we test for apis that have that kind of callbacks is that the callback also runs with expected context. To do this we add a span named parent around the operation we are going to test and create a span named child in the callback. Span from callback should be a sibling for the span for the executed operation. See

opentelemetry-java-instrumentation/instrumentation/okhttp/okhttp-3.0/testing/src/main/java/io/opentelemetry/instrumentation/okhttp/v3_0/AbstractOkHttp3Test.java

Line 148 in 42ba962

void requestWithCustomInterceptor() throws Throwable {

for an example, you can find more samples by searching for "child".

[steverao]

One thing we test for apis that have that kind of callbacks is that the callback also runs with expected context. To do this we add a span named parent around the operation we are going to test and create a span named child in the callback. Span from callback should be a sibling for the span for the executed operation. See

opentelemetry-java-instrumentation/instrumentation/okhttp/okhttp-3.0/testing/src/main/java/io/opentelemetry/instrumentation/okhttp/v3_0/AbstractOkHttp3Test.java

Line 148 in 42ba962

void requestWithCustomInterceptor() throws Throwable {

for an example, you can find more samples by searching for "child".

I added relevant callback tests. But I found because of the implement of callback logic in https://github.com/influxdata/influxdb-java/blob/0c690c8d8a6627f525e05d68e0571e089c61845e/src/main/java/org/influxdb/impl/InfluxDBImpl.java#L637 the child span is the child of the executed operation instead of sibling:(

[laurit]

I implemented running callbacks in the scope of the parent span.

[laurit]

The naming rules for database client span names in https://github.com/open-telemetry/semantic-conventions/blob/main/docs/database/database-spans.md do not suggest using unknown when database name is not known, they suggest to omit the database name instead.

[steverao]

The naming rules for database client span names in https://github.com/open-telemetry/semantic-conventions/blob/main/docs/database/database-spans.md do not suggest using unknown when database name is not known, they suggest to omit the database name instead.

Thank you for your reminder, I modified it now. But I found in https://github.com/open-telemetry/semantic-conventions/blob/main/docs/database/database-spans.md, it is not clearly stated that it is recommended to omit the database name in this case. Is it necessary to add this situation in the document?

[laurit]

I reread this part in the spec and noticed that it was changed a couple of days ago. Now the spec suggests a prioritized list of attributes that could be added to the span name. We have not implemented this yet.

[steverao]

Repen to trigger CI tasks.