Postgres monitoring that helps you be proactive

From the team behind pgMustard

€50 per database per month, cancel any time.

A pgCustard dashboard. Highlights show a long-running transaction holding back cleanup for 2 days 4 hours, with an explanation of the impact and options to fix it, alongside two indexes with over 95% estimated bloat. Below, database checks such as version, uptime and cache hit ratio are marked green, with xmin horizon blocked in red and autovacuum workers in yellow.

Comprehensive Postgres monitoring can be expensive, and/or complex to set up. As such, smaller teams are often missing important checks, and currently firefight issues as they arise.

By checking certain things proactively, problems can be spotted and fixed before they become acute. This allows them to be mitigated during normal business hours, preventing both stress and incidents. Yay!

On the cost side, managed service providers often double their prices at each step, which can get painful. By monitoring your top queries, and checking their potential to be optimised, you can extend the time between upgrades. In some cases, downgrading is even possible!

If you imagine what a monitoring tool built by the team behind pgMustard would be like, you’d get pgCustard. Simple, actionable, and nuanced advice, that doesn’t assume you’re either an expert or a novice.

A pgCustard issue page for 96% estimated index bloat. It summarises the index’s size, estimated bloat and scans, suggests REINDEX INDEX CONCURRENTLY, and explains how the estimate works.

If a critical issue appears, we’ll email you. Each week, we also email you a summary, so you can spot trends before they become issues.

A pgCustard weekly summary email. A “This week” table compares the database with the week before: size up 31GB to 2.1TB, mean query time up 0.1ms to 1.6ms, and a longest cleanup hold of 2d 4h, among others. Below it, the largest tables start with public.page_views at 612GB.

It'll likely be ideal if you're

  • A small team without a DBA
  • Excellent engineers who aren’t Postgres experts
  • Already having occasional issues

What does pgCustard monitor?

  • Major and minor version
  • Table vacuum pressure
  • Table analyze pressure
  • Table bloat
  • Index bloat
  • Unused indexes
  • Redundant indexes
  • Invalid indexes
  • Xmin horizon blockers
  • WAL-forced checkpoints
  • Deadlocks
  • Autovacuum workers
  • Non-idle sessions
  • Cache-hit ratio
  • Mean execution time (overall, and for your top queries)
  • Database size
  • And a few more!

What are the requirements?

  • pg_monitor or pg_read_all_stats role
  • PostgreSQL 14–19
  • pg_stat_statements (optional)

Which providers are supported?

We’ve tested on each of Amazon RDS and Aurora, Google Cloud SQL and AlloyDB, Azure, Aiven, ClickHouse, Crunchy Bridge, DigitalOcean, Heroku, Neon, PlanetScale, Render, Supabase, Tiger Cloud, and Xata. Other services will likely work too, and so does monitoring databases you host yourselves.

What about sharded or distributed databases?

We haven’t tested sharded setups like Citus or Neki, or distributed databases like CockroachDB or Aurora DSQL, so they probably won’t work well yet. If you’d like us to support one, please let us know.

Does my database need to be reachable from the internet?

No. You can run our open-source collector inside your own network to send the metrics to us. Otherwise, we connect over the internet from our static IP address, which you can allowlist.

What’s the overhead?

Very low, especially if you have pg_stat_statements on already. pgCustard polls every 30s over a single connection, running about 10 queries, which take around 30ms in total on a database with 100 tables.

What data do you collect, and how do you protect it?

Statistics and metadata, such as table names, sizes, and normalised query text. See our privacy policy and our security page for details.

Are the recommendations LLM generated?

No. Much like our pgMustard product, we’re sticking to human-written advice based on deterministic calculations. This makes it easier for us to keep advice safe, to debug issues, and to improve things over time.

Can I sign up straight away?

Not yet. We’re opening up gradually, and expect self-service sign-ups to open in December 2026. Until then, you can request early access.

How much does early access cost?

€50 per database per month. Early access also gets you a shared Slack channel (or similar) with us, where we can advise further on the issues raised by pgCustard.

Is there a free trial?

No, but pgCustard should be valuable straight away. If it isn’t, let us know within 30 days of your first payment, and we’ll refund it in full.

Does it integrate with pgMustard?

No. At least for now, the products are totally separate.

Was pgCustard originally an April fools joke?

Yes it was, but this time we’re serious!

€50 per database per month, cancel any time.