V1 [OBSERVE] Error breakdown by HTTP status and type

[austin-barrington]

Context

The current metrics surface total errors (hyperbytedb_errors_total) but provide no breakdown. Operators cannot distinguish between:

• Client errors (4xx: bad queries, auth failures, invalid line protocol)
• Server errors (5xx: internal failures, chDB errors, WAL write failures)
• Specific error types (rate limit exceeded, auth denied, query parse failure, chDB execution error)

Deliverables

• Replace or supplement hyperbytedb_errors_total with hyperbytedb_http_requests_total{status="2xx|4xx|5xx",method="POST|GET|DELETE",endpoint="/write|/query|..."}.

• Replace hyperbytedb_query_errors_total and hyperbytedb_write_errors_total with more specific counters:

  • hyperbytedb_wal_write_errors_total
  • hyperbytedb_chdb_query_errors_total
  • hyperbytedb_chdb_write_errors_total
  • hyperbytedb_auth_errors_total{reason="invalid_password|user_not_found|token_expired"}

• Add hyperbytedb_query_parse_errors_total to track malformed queries separately from execution errors.

• Update docs/user-guide/monitoring.md with the new metric names and recommended alert thresholds.

References

• HTTP handler error responses — src/adapters/http/
• Query and write services — src/application/
• docs/user-guide/monitoring.md — existing metrics docs

Priority

P2 — Observability, operations