Pglocks.org

80•hnasr•1y ago

Comments

whilenot-dev•1y ago

I'm a bit lost here.

Locking is a challenging problem in complex systems. Is this list to be interpreted as a "TODO: get rid of locking conflicts in future releases" or more a "NOTE: be aware there are known conflicts that will not change - find ways to work around them"?

EDIT: Also, is the creation of this list an automated or a manual effort?

tux3•1y ago

I think this is intended as educational material, not a list of things to fix.

The locks are here by necessity, it is not so easy at all to get rid of them. And even in special cases where it is possible, the complexity you have to introduce is not to be taken lightly...

If even a tenth of these disapppeared, it would be incredible, in a very surprising way.

atombender•1y ago

The creator looks like a developer and teacher, not a Postgres core team member. So I assume this is for documentation purposes.

I actually like this a lot, as there isn't a single place in the Postgres documentation that lists all the possible locks; it's spread out all over. Having a quick reference for what kinds of commands you'd be blocking with your transaction is valuable.

It's pretty evident that the pages have been programmatically generated, but I'd love know what it's generated from. I think you can derive this information from the documentation, but not sure if you can do it in an automated way without an LLM.

braiamp•1y ago

> there isn't a single place in the Postgres documentation that lists all the possible locks

Did you read this page? https://www.postgresql.org/docs/current/explicit-locking.htm...

atombender•1y ago

That's a great page, but it has several issues.

First, it isn't complete; as I said, the locking behaviour is spread out all over the Postgres documentation. For example, that page doesn't list what locks DROP INDEX takes. To find that out, you have to go to the documentation page for that command and read it carefully. In fact, really carefully — the locking behaviour is only documented under the section about CONCURRENTLY.

The page also doesn't list what possible commands are then blocked. Locks interact in subtle (and incorrectly named!) ways that are explained in the tables on that page ("Conflicting lock modes"), so to understand if something will block something else you have to look at the two commands you are curious about and then look at how their locks interact.

gulcin_xata•1y ago

I agree, it is not so straightforward to find out.

braiamp•1y ago

These are database locks, which means that depending which arrives first, the later transaction has to wait till the first one finishes to complete. These locks are about SQL commands and which commands can run concurrently with the others. There's a graph here of how that looks like https://pankrat.github.io/2015/django-migrations-without-dow...

Usually for maximum performance (minimum latency, maximum throughput) you want to have operations not lock each other, unless absolutely necessary, in which case you want them to be short.

whilenot-dev•1y ago

You make it sound like the conflict is just affecting performance and won't result in a deadlock. So it's for performance aware postgres clients/users, and not for postgres developers?

andyferris•1y ago

It is a guide for developers using postgres as a client, who need to write systems that don't deadlock, are performant and are correct. These are the (rather sharp) tools that postgres provides for doing so (or else you can use e.g. serializable isolation and optimistic concurrency, but in my experience that has too many false positives and bail out rather eagerly, whereas these tools let you be very precise and granular).

mebcitto•1y ago

Other relevant talks/blogs that I found really useful for understanding Postgres locks are:

* Unlocking the Postgres Lock Manager by Bruce Momjian: https://momjian.us/main/writings/pgsql/locking.pdf

* Anatomy of table-level locks by Gulcin Yildirim Jelinek: https://xata.io/blog/anatomy-of-locks

pasxizeis•1y ago

Shameless plug: I wrote a tool[1] that executes a given migration against a test database (e.g. in your CI) and reports back what locks it acquired.

The rationale being to have a "lock diagnostics report" commented in your PR's migration file.

It's a prototype and has a few rough edges and missing functionality, but feedback is more than welcome.

[1] https://github.com/agis/pglockanalyze

jononor•1y ago

Very practical! Locking is one of the things that can really bite when doing migrations.

USPS now offers official ZIP Code boundary polygons

TPUs

Claude Fable 5.2 Scores 65 on Artificial Analysis, Highest So Far

Google Gemini routes personal Drive files and emails to human annotators

Tracking the Harm of DOGE Cuts

E-means: 10x faster k-means at equal loss

Payments Canada announces launch of Canada's Real-Time Rail (RTR) payment system

Our Smallest, Most Efficient Model: Muse Ember

CUI – Native UI components for C

Fold Epic Paper Airplanes

ssh2RCE

Infineon marks innovation milestone with 300mm power silicon carbide wafer

Magenta solar panels may be agrivoltaics' cleverest trick yet

Parking-app convenience fees in 312 U.S. cities (median $0.35 per session)

Claude Fable 5.2 and Mythos 5.2

A primer on GPUs and kernel optimization

Show HN: Every satellite that passed over any place this week, and its photos

Exact – a new cross platform UI framework by the Expo team

Is Opus 5.5 nerfed? Thinking patterns changed on November 3

Y/N (YC S23) is hiring a growth lead to distribute superhuman A.I. storytelling

Show HN: MeowDisplay Is iPhone Mirroring, but in Reverse

An ode to livelyness analysis; SSA to registers

Magnets and Rare Earths

Jensen Huang Scandal Sparks Speculation–Jensen's Wife Could Get Half

Possible 19th-century castaway camp found on Isla de los Estados

Deuterium's Impact on OLED Displays (2024)

What is malloc doing, anyway?

Show HN: Someone Was Here – Find out who used your Mac while you were away

DeepX claims inference up to 100B parameters at less than 5 W

Show HN: DBDelve – Open-Source DB Client Made with Rust and GPUI