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.
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:
/+/users/+/logins/<tenant>/users/<userid>/logins/+/sensors/+/<tenant>/sensors/<sensorId>/+/comments/+/<tenant>/comments/<threadid>/+/tls-certificates/+/<tenant>/tls-certificates/<managedCertificateId>Also we can get:
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.