pgconfigurator
pgconfigurator
pgconfigurator · v2 · by CYBERTEC

Tune PostgreSQL.
Decode every plan.

A modern replacement for the pgconfigurator you know — paired with a plan analyzer that tells you what's wrong, not just what's slow. Free, fast, and in your browser — your data never leaves.

Paste anything — we'll figure it out

EXPLAIN output, DDL, postgresql.conf, SHOW ALL — paste any PostgreSQL artifact and we'll route it to the right tool.

parsed in your browser · nothing uploaded
no signup · no tracking by defaultor openTuneExplainInspectdirectly
pgconfigurator.cybertec.at
TuneExplainInspect
Tune · OLTP · PG 17
postgresql.conf
5 changes
shared_buffers
128 MB8 GBlive reload
work_mem
4 MB32 MBlive reload
effective_cache_size
4 GB24 GBlive reload
max_connections
100300restart
checkpoint_timeout
5 min15 minlive reload
32 GB RAM · 8 vCPU · NVMeclick any row → why
Explain · plan #4f2a
orders ⨝ users
4.2 s total
Aggregatecost=1.2M rows=1 time=4218ms
└─Gatherworkers=4 rows=8.4M time=4195ms
└─Hash Joinrows=2.1M time=4011ms
└─Seq Scan on ordersrows=4.0M time=2890ms
└─Hash → Index Scan usersrows=180k time=86ms
Seq Scan over 4M rows · 69% of total time
Likely fixable with CREATE INDEX ON orders(user_id, created_at). Expected speedup: ~14×.
Tune · workload: OLTP · 32 GB RAM · PG 17
postgresql.conf
shared_buffers128 MB8 GB
work_mem4 MB32 MB
maintenance_work_mem64 MB1 GB
effective_cache_size4 GB24 GB
wal_buffers−164 MB
checkpoint_timeout5 min15 min
memory · cache
shared_buffers
safe live reload
default
128 MB
suggested
8 GB
your override
Why this value: with 32 GB of RAM dedicated to PostgreSQL, the convention is 25% for shared_buffers. Going higher rarely helps because the OS page cache already serves the same data; going lower starves frequently-hit pages.
Related
effective_cache_size — should be ~3× of this
huge_pages — recommended on Linux for this size
docs · PG 17 · runtime/resource
Tune

Every value, justified.

A configuration you can defend in a review. Every parameter ships with the reason it was chosen, its risk class, and the context to know when to deviate — so the file you hand off is one you understand line by line.

  • Click any parameter to see why we suggest that value, what it controls, what happens if you get it wrong, and which docs section to read next.
  • Every change is risk-graded — safe live reload, requires restart, or proceed with caution — so ops people know exactly what they're signing up for.
  • Outputs match how you deploy: postgresql.conf, ALTER SYSTEM, Ansible task, Patroni-flavored YAML, or a Markdown runbook.
Explain

Plans that tell you what to do.

Reading a plan is the easy part — knowing what to change is the job. The findings panel runs the checks experienced DBAs run by hand, every time, and writes the SQL you'd type next.

  • Multiple synchronized views of the same plan: a tree for structure, a timeline for what each parallel worker actually did, and a jitter view that surfaces where runtime varies between executions.
  • Findings are the whole point. Anti-pattern detection flags bad estimates, lossy bitmap scans, on-disk sorts, hash spills — with the exact fix to apply.
  • Paste two plans, see the diff. Verify that your index actually helped, your config change actually paid off, your upgrade actually shipped what the changelog said.
Privacy

Your plans never leave your browser.

An execution plan carries more than the query: table and column names, literal values from your WHERE clauses, row counts that reveal how much data you hold, index and partition names, even timings that hint at your hardware. That's exactly why pgconfigurator parses and analyzes plans entirely in your browser — nothing is uploaded.

no data sent · no tracking · no third-party scripts
Parsed in your browser
Plan analysis and config logic run as WebAssembly. The default path never sends your data anywhere.
EU-hosted, transient server
Our optional fallback endpoint runs on EU infrastructure. Requests are processed in memory and not stored.
PII sniffer built in
Pastes containing likely personal data are flagged. One-click deterministic anonymization preserves readability.
GDPR by design
No login, no tracking scripts, no third-party fonts. Privacy policy reads in a single page.
Behind the tool

Built by people who teach PostgreSQL.

CYBERTEC has spent a quarter-century inside other people's PostgreSQL — tuning it, scaling it, recovering it at 3 AM, migrating onto it from Oracle and SQL Server, and contributing back upstream. Every heuristic in this tool comes from that pile of experience, written down.

25+
years of PostgreSQL
8
PostgreSQL committers & contributors
EU
engineering, hosting, support
24/7
incident response for customers
“The old pgconfigurator told us what. The new one tells us why. It saves us a week of writing internal docs every time a new engineer joins.”
DBA team lead
Financial services · EU
Talk to humans

Need more than a tool?

The configurator gets you 80% of the way. For the rest — an audit before you ship that migration, an HA cluster that won't wake you up, a 24/7 support contract — our engineers are one form away.

Performance audits and emergency tuning
HA architectures, Patroni, replication
Major version upgrades, zero-downtime
Migrations from Oracle, SQL Server, MySQL

Fields marked * are required.