Postgres Decoded
Each episode takes one PostgreSQL concept — a query planner decision, a locking scenario, a replication edge case — and examines it without rushing through the details.
Recent Episodes
Text transcripts of each session are archived in the archive section.
-
38
Logical Replication Slots and What Happens When They Stall
Stalled replication slots accumulate WAL on disk silently. Most operators discover this problem only when disk space is already critical. The episode walks through monitoring queries and safe slot termination procedures.
-
37
Index-Only Scans: When They Work and When They Quietly Fall Back
The visibility map controls whether Postgres can skip heap fetches entirely. High update frequency on a table can make index-only scans nearly useless despite ideal index coverage.
-
36
Row-Level Security Patterns for Multi-Tenant Schemas
RLS policies introduce a measurable overhead depending on how they are written. Joining to a session variable set at connection time differs substantially from embedding subqueries inside policy definitions.
-
35
VACUUM Tuning for Tables That Receive Constant Writes
Autovacuum thresholds calculated from table row count become impractical on tables exceeding 200 million rows. Switching to cost-based throttling with per-table overrides is the documented approach for these cases.
-
34
Generated Columns vs. Triggers for Computed Fields
Generated columns in Postgres 12+ are stored and updated automatically but carry restrictions: they cannot reference other generated columns or call volatile functions. Triggers offer flexibility at the cost of explicit maintenance.
What the series covers
Episodes are grouped loosely around operational areas. Each session assumes familiarity with SQL and basic database administration concepts — there is no introductory material repeated across episodes.
Guests include database engineers from companies running Postgres at scale, and contributors who worked on specific features discussed in the episode.
Query Planning
Plan nodes, cost estimation, statistics, planner hints.
Concurrency
MVCC internals, lock modes, deadlock analysis.
Storage & I/O
Tablespaces, partitioning, TOAST, write amplification.
Extensions
pg_stat_statements, pgvector, TimescaleDB, PostGIS use cases.
Episodes published since 2014
Topic areas in rotation
Postgres-only focus, no generic SQL theory
English, accessible to international listeners
How each episode is structured
Sessions open with a concrete scenario — a slow query, a replication gap, an unexpected lock — and spend the majority of runtime examining the mechanism behind it. Code samples and configuration excerpts are read aloud and published in full as text transcripts.
Questions from the Domain seminar community are answered in the final segment of most episodes. Submit questions via [email protected] or through the contact page.