Welcome to the March 2023 edition of the pganalyze newsletter!
Today, we share the recording of our recent webinar "How to use the Postgres query planner to debug bad plans and speed up queries". It was exciting to see that over 430 of you joined us in our live session! As always, it was great to share Postgres best practices and to answer everyone's questions. We're planning our next webinar for June, so keep an eye out on the newsletter!
On the product side, we've continued testing some of the new autovacuum and bloat estimation features, as well as worked on new tuning recommendations for per-table autovacuum settings, soon available for early access. If you're interested to try this out, let us know!
You can also find our 5mins of Postgres episodes further below in the newsletter, where we've talked about topics like pg_stat_io, handling integer sequence overflow, UUIDs vs serial for primary keys, and much more.
All the best,
Lukas
Postgres Planner webinar recording
We're proud that over 430 people joined the live session last week. In case you missed it or want to watch the recording, check out the video below! We talked about how the Postgres query planner chooses different plans for the same query, and how to debug bad plans when they occur.
In detail, we discussed how to use auto_explain, how to identify a bad plan, how to get Postgres to switch plans to improve performance, and more.
Over 140 of you took the time to give us feedback on our post-webinar survey. We appreciate you taking the time to do so and will work on implementing it to make our webinars and pganalyze more useful to you!
We're continuing to work with existing pganalyze customers on testing the new autovacuum statistics and bloat detection logic.
If you're interested to join the early access program and get access to the new functionality, just register and we'll reach out with next steps (you may have to upgrade the collector to the latest release, amongst other details).
We're also gearing up to share our new per-table autovacuum tuning recommendations soon (ever wondered whether you should tune settings like scale factor and threshold for certain tables?), and registering now gets you a spot to be amongst the first to get access.
Learn reasons for using random keys and for using sequential keys. We look at hybrid approaches (sequence + random or time + random) and are looking at UUIDv7.
PgCat is a Rust-based connection pooler that tries to be compatible with pgbouncer. It adds features like load balancing as well as some sharding options.
We show what collations are and specifically look at glibc and ICU collations. We flag problems that can occur with collations when using different OS releases.
Subscribe on YouTube
We publish weekly videos on Postgres, our conference talks, tutorials, product walk-throughs, and more.
Support for new Amazon RDS and Aurora SSL certificates
This upgrades the included certificate set to the new global CA certificates, which includes "rds-ca-2019-root" and all newer RDS CAs such as "rds-ca-rsa2048-g1"
Better handling of locked relations during schema data collection
This fixes a race condition that may have missed locked tables to be reported as such (even though their statistics were ignored to avoid blocking on the lock), causing missing data in pganalyze
Additionally this fixes a related issue that could have caused bogus spikes in table statistics when the table was locked
As well as a handful of other bugfixes
Index Advisor and Query Analysis
Fix handling of generic query plans that produce AlternateSubPlan nodes
Type width estimates: Avoid errors for certain variable-length types provided by extensions (geometry, citext)
VACUUM monitoring & tuning
We've continued iterating on new functionality available in early access (see above)
Other improvements
Alert policy configuration: Remove "Coming soon" flag from blocking queries check
Automated EXPLAIN: Prefer storing samples with EXPLAIN plans when too many samples are received within the same timeframe
This resolves some edge cases where a query sample was visible, but no EXPLAIN plan, despite a plan having been collected
Review us on G2
Please leave your feedback (good or bad!) in this survey - we appreciate it!
