Logging, health, and diagnostics

Log levels, trace fields, startup banner, and diagnostic commands.

Version
Latest
v0.1.0 · latest 2 min read
On this page
  1. Levels
  2. Trace fields
  3. Banner
  4. Commands
  5. Metrics SQL surface

Levels#

Default info. PLOMID_LOG_LEVEL=debug ./start.sh or RUST_LOG=server=debug,server::protocol=trace ./start.sh. Use info operations, debug lifecycle, trace protocol/parser. RUST_LOG=plomid::perf=debug for per-op timers.

Trace fields#

Every connection/query log: connection_id query_id statement_type sql_len elapsed_ms (+ error context). Values/passwords excluded.

PLOMID ENTERPRISE v<version>, PG wire v3, OS/ARCH, boot checklist, listen/port/data dir/role/auth (tls off|enabled(scram-sha-256|md5|password|disabled))/tls.

Commands#

shsource
./target/debug/plomid db readiness --port 5432
./target/debug/plomid db health --port 5432 --json
./target/debug/plomid db diagnostics --port 5432
./target/debug/plomid metrics --port 5432

live_connections gauge + drain(5 s) on shutdown; shutdown_checkpoint logged; maintenance worker shutdown reports {submitted, completed, inline_fallbacks}.

Metrics SQL surface#

In-process counters, exposed in-band (no ports, no scraping sidecar). All counters are process-wide and monotonic; histograms use fixed ≤1ms … >10s buckets with count/total/max (mean = total / count).

Function Families
plomid_stat_counters() queries, errors, connections, slow queries, maintenance, checkpoints, stale tables, TLS (tls_connections, tls_handshake_failures, tls_plaintext_refusals), analytical totals
plomid_stat_tables() per-table writes + is_stale freshness (only tables written since boot appear)
plomid_vector_indexes() per-index vector index state (builds, readiness, sizes)
plomid_stat_transactions() begun/committed/aborted/conflicts, active gauge, lifetime + commit-latency histograms
plomid_stat_wal() bytes/appends/covered, backlog gauge, rotations, flush + sync durations
plomid_stat_contention() engine-mutex, write-gate, group-commit barrier + formation, maintenance-queue waits
plomid_stat_buffer() pool hits/misses/evictions, last-sync dirty sample, checkpoint bytes + durations
plomid_stat_columnar() segments examined/skipped, columns read, batches, spill bytes/files, stale tables
plomid_stat_analytical() fast-path counts by path
plomid_htap_stats() admission, memory hierarchy, result buffers
plomid_vector_stats() exact/ANN/sparse/binary, cache hits/misses, fallbacks, builds, rebuilds + duration, fresh/stale
plomid_vector_backend() compute setting, CPU arch/threads, GPU requested vs available
plomid_stat_security() 7 counters: auth_failures, authz_denials, account_lockouts, legacy_auth_uses, tls_connections, tls_handshake_failures, tls_plaintext_refusals
plomid_stat_admin() admin actions by name (reindex, vacuum)

When stuck: Troubleshooting.

Was this page helpful?