Skip to content

RELEASE_NOTES.md

Latest commit

 

History

History
91 lines (64 loc) · 4.19 KB

RELEASE_NOTES.md

File metadata and controls

91 lines (64 loc) · 4.19 KB

postgres_dba 7.0

34 reports | Tested on PostgreSQL 13–18 | Works with pg_monitor role

New Reports

Corruption checks (c1, c2, c3) — powered by amcheck

Three levels of integrity checking, all requiring CREATE EXTENSION amcheck:

Report Lock What it checks When to use
c1 AccessShareLock B-tree pages, GIN indexes (PG18+) Production — fast, safe, non-blocking
c2 AccessShareLock c1 + heap/TOAST integrity (PG14+) Production — safe but reads all data
c3 ShareLock ⚠️ B-tree parent-child ordering, sibling pointers, rootdescend, checkunique (PG14+) Clones or standbys — detects glibc/collation corruption
c4 ShareLock ⚠️⚠️ Everything in c3 + heapallindexed + verify_heapam with full TOAST Clones only — proves every heap tuple is indexed, slow on large DBs

All three check system catalog indexes (pg_catalog, pg_toast) — because catalog corruption is the scariest kind.

Robustness:

  • Graceful handling when amcheck extension is not installed
  • No false corruption reports on insufficient privileges (reports skipped count)
  • Version-conditional: uses appropriate function signatures for PG11–18
  • GIN support via gin_index_check() on PostgreSQL 18+

snapshot.sh — LLM-friendly output

New snapshot.sh script dumps all safe reports as clean plain text — no ANSI colors, no interactive prompts, no psql noise. Perfect for:

  • Feeding database state to an LLM for analysis
  • Automated health checks in scripts
  • Saving periodic snapshots to files
./snapshot.sh -d mydb > snapshot.txt
./snapshot.sh --full -d mydb    # include expensive reports too

m1 — Buffer cache contents

What's in your shared_buffers right now. Shows cached size vs total size, % of cache used per object, and dirty buffer counts. Requires pg_buffercache extension.

s3 — Workload profile by query type

Groups pg_stat_statements by first SQL keyword (SELECT, INSERT, UPDATE, DELETE, etc.) to show workload composition at a glance. Correctly handles queries with leading block comments (/* ... */) and line comments (-- ...).

t2 — Objects with custom storage parameters

Lists all tables, indexes, and materialized views with non-default reloptions. Flags potentially problematic settings: disabled autovacuum on large tables, low fillfactor, aggressive vacuum scale factors. Shows partition relationships.

Report 0 — WAL and replication slot info

The node information report now includes:

  • WAL: current LSN, file count, total WAL size
  • Replication Slots: name, type, active/inactive status, lag from current WAL position

Report Renames

Categories reorganized for consistency:

Old New Reason
b6 m1 Buffer cache → m (memory) category
c1 p1 Index creation progress → p (progress) category
p1 x1 Alignment padding (experimental) → x (experimental) category

Bug Fixes

  • i3: Fixed operator is not unique error when intarray extension is installed (added explicit ::int2[] cast)
  • s3: Fixed function round(double precision, integer) does not exist — added ::numeric casts
  • i2: Removed unused redundant_indexes_grouped CTE (dead code)
  • s1: Removed duplicate sum(calls) in the pre-PG13 code path

Terminology

  • Updated Master → Primary across all reports and CI (0_node, i2, i4, i5)

Typo Fixes

  • inspect → inspects (b1, b2)
  • filed → fields, fractionnal → fractional, functionnal → functional (b2)
  • alt_shits → alt_shifts (p1) 🙈
  • Vaccuum → Vacuum (b1)
  • format → formatting (s2)
  • Comment formatting: added space after -- throughout (b3, b4, l1, s2, v2)

CI Improvements

  • Added PAGER=cat to all psql invocations (prevents pager hangs)
  • Added intarray, pg_buffercache, and amcheck extensions to test matrix
  • Added foreign key test tables for i3 regression testing
  • Added dedicated i3 regression test with intarray installed

Compatibility

Tested on PostgreSQL 13, 14, 15, 16, 17, and 18 — all 34 reports pass with both superuser and pg_monitor roles.