confidential-containers / trustee

Attestation and Secret Delivery Components
Apache License 2.0
64 stars 83 forks source link

Add metrics endpoint #543

Open fitzthum opened 1 day ago

fitzthum commented 1 day ago

Currently the only way to figure out the status of Trustee is to look at the logs. We should add a more sophisticated metrics infrastructure that can be connected into third-party systems.

mkulke commented 20 hours ago

Note that while prometheus support is a good idea, it will not supplement logs. metrics are about aggregate values (think: number of successful tdx verifications per minute), you shouldn't/cannot debug individual transactions with it (because of restrictions on the cardinality of metrics).

the closest to a supplement logs today are opentelemetry traces. sadly, this is still a bit rough for rust webservices today. but embracing structured logs via tracing is a step in the right direction. e.g. there should be correlation ids to debug logical transactions (like an RCAR ceremony).