PostgreSQL 9.0b1 with built-in replication support

http://developer.postgresql.org/pgdocs/postgres/release-9-0.html

by andrevoget • 16 years ago

135 19 16 years ago

developer.postgresql.org

astine2• 16 years ago

Replication support has been a big issue with Postgres in the past, with folks needing third party proxies that were clumsy to use to accomplish it. This improvement is big news.

It sounds like there are a lot of improvements with this release of Postgres, and that they are taking a less conservative approach to development than they have in the past; I hope that they won't sacrificed stability too much to accomplish this.

crad2• 16 years ago

I'm not sure what you base the less conservative approach to development comment on. As far as I know, the development process and patch application process is the same as it has been in recent years. Stability and quality of code are still #1.

Edit: clarified the conservative approach to development.

mahmud• 16 years ago

Perhaps the submission itself? From the first paragraph:

"While past major releases have been quite conservative in their scope, this release shows a bold new desire to provide facilities that every new and existing user of PostgreSQL will embrace."

a2tech2• 16 years ago

As with any major postgres update this calls from a full dump of the database and a restore into the new version-you can't simply update the database files.

pg_dump -clean makes this a snap so you shouldn't have any problems upgrading. (oh and if you're doing backups-you are using pg_dump --clean right? Its really the best way).

russss2• 16 years ago

If you use "pg_dump -F c", pg_restore is able to parallelize the restore, which makes things go a lot quicker.

a2tech2• 16 years ago

Wow-thats a great tip. Thanks!

pilif2• 16 years ago

maybe we'll get an updated pg_migrator once the release is out. While dump/restore is for sure the safer option, once your databases reach Terabytes in size, doing the migration in-place (most of the time the on-disc structure is quite similar between releases) will be a huge speed advantage.

defen• 16 years ago

Does this cause any problems if you have foreign key constraints? I remember running into that problem a couple years ago when restoring a pg_dump.

toisanji• 16 years ago

The best open source database just got better, can't wait to use this in production.

pilif2• 16 years ago

If I had one feature wish free, it would be the option of having an option to sort by multiple locale specific sorting orders in the same database.

Earlier, this was specified at cluster initialization time (initdb), then it was one a database level (create database), but what I would really like to see is being able to specify it on query time or, failing that, I would at least like to be able to create multiple indexes with different locales and use those to sort the data.

In the application in question I have to do all sorting on the client side now which also means that pagination needs to be done on the client side which is a huge networking overhead between database and application as I'm fetching rows, sorting them and then throwing away the majority of them.

Aside of that, I love, love, love PostgreSQL ever since I switched from MySQL back in 2000. Every new release was such a pleasure to try out and every time I was happily integrating all newly gained features in my applications.

eds• 16 years ago

Have you tried adding a function that converts between locales and ordering/indexing by that in Postgres?

Like in this thread: http://archives.postgresql.org/pgsql-hackers/2004-09/msg0075...

russss2• 16 years ago

The pg_collkey extension gives you access to ICU's unicode collation functionality from within Postgres:

http://www.flexiguided.de/publications.pgcollkey.en.html

Also see my blog post about it here:

http://russ.garrett.co.uk/2009/01/18/unicode-postgres/

tibbon• 16 years ago

I'm really excited to see this. Using 3rd party tools for PostgreSQL was getting confusing for me and having it built in will really help a ton.

Ionic_Walrus• 16 years ago

really excited about this, having dealt with slony-1 built-in replication is a welcome feature.

mark_l_watson• 16 years ago

Nice. Say goodbye to Slony? I am downloading the beta right now for play use. I would like an easier AWS setup with two EC2s running replicated PostgreSQL.

xal• 16 years ago

Is the build in replication support able to correctly replicate schema changes to the slaves? That was the biggest issue with the existing solutions for most admins I talked to.

thristian• 16 years ago

It's based on shipping write-ahead-logs around, so anything you can do inside a transaction (pretty much any DML or DDL statements except for ones that affect the entire cluster, like "CREATE USER" or "DROP DATABASE") can be replicated.

PostgreSQL 9.0b1 with built-in replication support

Related Stories