Proposal: Integrating OpenTelemetry for efficient troubleshooting and observability
- Dominant language
- C++
- Stars
- 3.6k
- Forks
- 468
- Avg merge
- 29m
- Merged PRs (30d)
- 1
Description
Troubleshooting and debugging has been a long pain for our system (e.g., #2311, #3190, #288, #725 )
Currently there are no unified and effective methods or mechanisms for logging and error reporting across the components in GraphScope (Flex). Maybe the CNCF project [OpenTelemetry](https://opentelemetry.io/) provides a viable solution for us:
- it supports many signals: including logs/metrics/**trace**. The **trace** seems very suitable in our situation, it carries info through many components in a complex system along with a request(e.g., a Cypher/Gremlin query)
- rich [instrumentation](https://opentelemetry.io/docs/instrumentation/) support, covering the languages GraphScope used.
- less instrumentation efforts, some SDKs even support automatic instrumentation.
- The signals are rich enough for monitoring the system status and are compatible/integratable to visualization systems.
Contributor guide
Assessment
This issue has not been assessed yet.