elasticsearch instrumentation creates too many span names, with unique document IDs.

[remram44]

Describe your environment
Using opentelemetry-instrumentation-elasticsearch 0.24b0 with elasticsearch 7.14.1

Steps to reproduce
Perform some index() and/or delete() calls with id parameter set

What is the expected behavior?
The number of span names created is reasonable and finite

What is the actual behavior?
An unbounded number of span names get created, containing the unique document names. This makes it hard to search through operations, for example in Jaeger:

The datamart.test.xxxx parts are the Elasticsearch document IDs.

Additional context
I have a patch for this incoming.

[owais]

According to the Otel spec, the operation name should be <db.operation> <db.name>.<db.sql.table>. Looks like we are capturing document names in the span op name which is more analogous to a row than a table. We should update it to <db.operation> <db.name>.<index.name> and never capture document name in the span operation name.

PRs welcome :)

[remram44]

The current instrumentation is way less sophisticated, basically using the request URL as the span name. I only dealt with the specific issue of document IDs in #705, changing it to the format you describe is a significant effort (probably would have to stop instrumenting perform_request() and instrument single methods instead).