Metrics by stream type #14
Description
The database itself could help us get universal timeseries metrics by counting encrypted (to exclude plaintext meta events like subscriptions etc.) log messages posted to stream patterns like:
| Countable thing | Stream pattern | Pattern meaning |
|---|---|---|
| User logins | /+/users/+/logins |
/<tenant>/users/<userid>/logins |
| Sensors observations | /+/sensors/+ |
/<tenant>/sensors/<sensorId> |
| Comments posted | /+/comments/+ |
/<tenant>/comments/<threadid> |
| SSL certificates renewed | /+/tls-certificates/+ |
/<tenant>/tls-certificates/<managedCertificateId> |
Also we can get:
- all in aggregate (over all tenants)
- breakdowns by tenant
This can be done somewhat securely without needing access to encryption keys if we only use metadata (log entry # per stream), assuming streams are granular enough.