pgMustard

@pgmustard.com

A visualisation tool for Postgres explain analyze that also gives performance advice. Posts by @michristofides.com‬. Site: https://pgmustard.com/

The upcoming Postgres 19 is set to include extra EXPLAIN info in several places: 🚦 An "IO" parameter (async info) 💡 Generated and supplied hint info (pg_plan_advice) ♻️ Memoize estimate fields 🪵 WAL full page image bytes We've added support for them all in @pgmustard.com 🎉

New blog post: Read efficiency issues in Postgres queries If you've got a query that's slowly degrading in performance, you might have a read efficiency issue. The root cause could be table bloat, index bloat, or data locality degradation. www.pgmustard.com/blog/read-ef...

Read efficiency issues in Postgres queries - pgMustard

A lot of the time in database land, our queries are I/O constrained. As such, performance work often involves reducing the number of page reads. Indexes are a prime example, but they don’t solve every...

pgmustard.com

Our Read Efficiency tips are now more efficient to read! 📝 Better wording, mostly for clarity 🔬 More specific to the scan type, and therefore shorter in most cases 🌟 Improved scoring, especially for Bitmap Heap Scans More details: www.pgmustard.com/changelog

A screenshot of two pgMustard windows, with an example query plan in each, and an arrow pointing from the old version to the new. Each has a Bitmap Heap Scan highlighted, and the Read Efficiency tip expanded.

Bad row count estimates are a common cause of slow queries. We've had a tip for them in pgMustard from the start, but we've just finished a revamp of it, mostly to make the advice in several common cases clearer. We also now report the ratios a little more naturally.

A screenshot of pgMustard, focused on a tip that reads: "Row Estimate: out by a factor of 120. The query planner estimated that this operation would produce 283,853 rows per iteration, but it actually produced 2,444 (out by a factor of 120). Bad row estimates can lead to suboptimal query plans. They can often be improved by running ANALYZE on the tables involved, by increasing column statistics, or by allowing Postgres to assess the correlation between columns (using multivariate statistics)."

It was a very small sample size, but we were impressed with the number of people running newer Postgres versions in our recent poll 🙌

A chart showing 48% on 17 or 18, 42% on 15 or 16, 10% on 13 or 14, and nobody (admitting to being) on 12 or older

We've revamped our "Operation on Disk" tips ✨ * Made them clearer * Made them more succinct (in most cases) * Improved the scoring * Mention hash_mem_multiplier (when relevant) * Show "Operation in Memory" in more cases, with the memory used * Updated the linked blog post

A side my side comparison highlighting mostly how much more succinct a new version of the tip can be. The tip wording of the new version is as follows:

This operation wrote out to disk to store temporary data.

Disk space used: 282 MB

Reducing the size of the operation would make it faster, especially if you can get it to fit into working memory.

You can do this by reducing the number of rows, with a LIMIT clause, or reducing the size of each row, by operating on fewer columns.

Another option is to increase the memory available, by adjusting work_mem.

Learn more about working memory (underlined, as it's a link)

Great write-up by Michael Christofides from @pgmustard.com www.pgmustard.com/blog/what-do... > It turns out that for the general case, boundary conditions and NULLs (of course!) always need to be considered, so you can get one or two extra Index Searches when these can’t be ruled out #postgresql

What do the new Index Searches lines in EXPLAIN mean? - pgMustard

In Postgres 18 you’ll now see “Index Searches” lines in EXPLAIN ANALYZE output. If like me you were wondering what those mean exactly, you’re in the right place.

pgmustard.com

New blog post: "What do the new Index Searches lines in EXPLAIN mean?" In Postgres 18 you'll start seeing things like "Index Searches: 1" on each of your index scans. We looked into what they mean, and how that information can be helpful. www.pgmustard.com/blog/what-do...

What do the new Index Searches lines in EXPLAIN mean? - pgMustard

In Postgres 18 you’ll now see “Index Searches” lines in EXPLAIN ANALYZE output. If like me you were wondering what those mean exactly, you’re in the right place.

pgmustard.com

The upcoming Postgres 18 is set to include some nice improvements to EXPLAIN: * BUFFERS on by default with ANALYZE 🎉 * Fractional "actual rows" (huge in some cases) * Quite a few new fields (including the very cool Index Searches) We've now added support for all of them. 💛