The slides from my talk at Percona Live 2017, A PostgreSQL Response to Uber, which I also gave as “Why Uber Was (Mostly) Wrong” at PGDay Nordic 2017, are now available.

The slides from my presentation, Corruption War Stories, are now available.

My slides from my PGConf US 2017 tutorial, PostgreSQL When It’s Not Your Job, are available now.

The slides for my talk Corruption War Stories are now available.

So, what’s weird about this plan, from a query on a partitioned table? (PostgreSQL 9.3, in this case.)

1
test=> explain select distinct id from orders where order_timestamp > '2016-05-01';
2
                                                                  QUERY PLAN
3
-----------------------------------------------------------------------------------------------------------------------------------------------
4
 HashAggregate (cost=347341.56..347343.56 rows=200 width=10)
5
   Group Key: orders.id
6
   -> Append (cost=0.00..337096.10 rows=4098183 width=10)
7
         -> 

The one thing that everyone knows about compositive indexes is: If you have an index on (A, B, C), it can’t be used for queries on (B) or (B, C) or (C), just (A), (A, B) or (A, B, C), right? I’ve said that multiple times in talks. It’s clearly true, right?

Well, no, it’s not. It’s one of

On a PostgreSQL primary / secondary pair, it’s very important to monitor replication lag. Increasing replication lag is often the first sign of trouble, such as a network issue, the secondary disconnecting for some reason (or for no reason at all, which does happen rarely), disk space issues, etc.

You can find all kinds of complex scripts that do

The slides for my talk, Securing PostgreSQL at PGConf EU 2016 are now available.

The slides for my presentation Unclogging the VACUUM at PGConf EU in Tallinn, Estonia are now available.

Everyone has their own style, unfortunately, on how they edit postgresql.conf. Some like to uncomment specific values and edit them at the point they appear in the default file, some like to tack overrides onto the end… and some do a mixture of those (don’t do that).

My personal preference is to leave everything in the default file commented,