Opentelemetry-specification: Consider health checking as part of specifications

Created on 30 May 2020  路  2Comments  路  Source: open-telemetry/opentelemetry-specification

next to tracing, logging and metrics, health checks can be considered as one of the corner stones in any monitoring system. maybe would be useful to also specify health checking in OpenTelemetry.

The scope could be:

  • definition of the API endpoint where to find what information
  • two basic status services for readiness/liveness
  • a service returning a listing of all executed health checks
  • data structures how a health record is supposed to look like
  • support to implement health checking in the various programming languages

e.g. within the Java ecosystem Spring implements health endpoints as part of there actuators (https://docs.spring.io/spring-boot/docs/current/reference/html/production-ready-features.html#production-ready-health). Could be a start.

needs discussion question after-ga

Most helpful comment

rather complementary. The tickets in question ask for a mechanism to exclude certain URLs from tracing. Most frequent use case are health-related endpoints. But how those health endpoints actually look like is still open. Everybody is doing his own thing. And when people have to integrate many components, they have to deal with that mix of health information (in the lucky case components event have such an endpoint). I think it would be rather useful to also have a standard for this in place. It will then allow to both ease integration and allow further tooling on top of that to analyze and visualize.

All 2 comments

I think your issue is a duplicate of at least one of the below:

  • #173 "Q: Way to ignore healthcheck traces when using automatic tracer across all languages?"
  • #623 "standardize on exclusion functionality for HTTP instrumentation"

rather complementary. The tickets in question ask for a mechanism to exclude certain URLs from tracing. Most frequent use case are health-related endpoints. But how those health endpoints actually look like is still open. Everybody is doing his own thing. And when people have to integrate many components, they have to deal with that mix of health information (in the lucky case components event have such an endpoint). I think it would be rather useful to also have a standard for this in place. It will then allow to both ease integration and allow further tooling on top of that to analyze and visualize.

Was this page helpful?
0 / 5 - 0 ratings