Waiting for Postgres 16: Cumulative I/O statistics with pg_stat_io

https://pganalyze.com/blog/pg-stat-io

by lfittl • 3 years ago

168 29 3 years ago

pganalyze.com

ggregoire• 3 years ago

Unrelated with the topic but pganalyze seems great (never heard about it before).

That's how you make a good home page.

lfittl2• 3 years ago

OP here - thanks for noting!

We've recently invested effort to make the homepage better (added the animation, revised the messaging, etc), so that makes me happy to hear & will forward to the team as well.

tommasoamici• 3 years ago

Hey I added your blog to my RSS feeder and it seems that not all articles from your site are in the feed. I can’t inspect feed.xml from mobile to confirm if it’s a problem on your end or mine though

trollied• 3 years ago

I see lots of similarities between the (newer) Postgres wait analysis logging/views and Oracle AWR/ASH. I suspect it was heavily inspired.

yjftsjthsd-h• 3 years ago

Inspired, or solving similar problems in a similar tool leading to similar solutions?

DBAOneDay• 3 years ago

Indeed, shouldn't patents be checked to ensure there isn't any risk here?

anarazel2• 3 years ago

Wait events are perhaps inspired by oracle's stuff, at least indirectly. It was developed by Robert Haas and Amit Kapila, working for EDB. Traditionally a lot of their customers have migrated from Oracle.

The redesigned way of storing stats in memory and pg_stat_io aren't inspired by Oracle. Melanie, who authored pg_stat_io, has never used oracle. I haven't used oracle in ~15 years, and the only thing I remember is fighting crashes in connect_by (I influenced the design of pg_stat_io, and authored large portions of the shared memory stats stuff).

Edit: noted Amit's co-authorship of wait events

dboreham• 3 years ago

Good to see this part of the performance problem space being addressed.

sgt2• 3 years ago

Was thinking of upgrading to 15. So should I wait for 16?

1500100900• 3 years ago

While waiting for PG 16 you can upgrade to 15.

sgt2• 3 years ago

Agreed. I also see PG 16 is scheduled for release later in the fall. Also, you don't want to put the first 16.0 version on a production system, rather wait until 16.1, 16.2 etc.

macspoofing• 3 years ago

Because PostgreSQL major version upgrades are hard and for large databases, requires some level of downtime. Given that that major version releases are on an annual release cycle, staying evergreen is really really annoying.

fswd• 3 years ago

Does anyone know if 16 will bring back JSON_TABLE from Postgres 15? They initially had it in the beta, then at the last moment, removed it.

anarazel2• 3 years ago

Not yet clear. A lot of the infrastructure needed to support JSON_TABLE et al in a cleaner fashion, has now been added.

darksaints3• 3 years ago

It would be nice if the Azure and AWS Postgres teams could get off their asses and release version 15 already. It has been out for several months now, and crickets.

Homebrew too, but at least they have the excuse that it is open source and unfunded.

redwood• 3 years ago

Stripminers have to rest too

prettygood• 3 years ago

Same goes for GCP. Normally they release it before Christmas but version 15 still hasn’t been released.

sa46• 3 years ago

Homebrew does have Postgres 15. https://formulae.brew.sh/formula/postgresql@15

darksaints3• 3 years ago

It’s keg only, which shouldn’t matter, except that it does if you want to add any extensions. For example, postgis can’t be added to Postgres 15 using normal brew commands (you can if you edit it and build from scratch and link manually, but that’s kind of a nightmare). Everything is set up to treat PG14 as the default, because that is the one that is not a keg.

Winsaucerer• 3 years ago

I realise it may not be possible for many people due to their organisational requirements, but I think self-hosting is a great option when possible, and the backup story is great with pgbackrest.

Easy to upgrade when you're ready.

brightball• 3 years ago

Just use Crunchydata. Runs in any of the clouds and gives you better options.

jasfi• 3 years ago

It can be a good thing to wait a minor version or two before upgrading to the latest major version. There's much more confidence in the release that way, some companies even have a policy along those lines.

darksaints3• 3 years ago

15.1 came out in November.

Waiting for Postgres 16: Cumulative I/O statistics with pg_stat_io

Related Stories