Architecture & Pitfalls of Logical Replication
Atsushi Torikoshi is a software architect at NTT OSS Center. He joined NTT group in 2007.
No video of the event yet, sorry!
We, at NTT OSS Center, provide NTT group companies with technical support service for open source software. During our experience of over 10 years providing technical support for PostgreSQL, customers have often demanded measures like replicating partial tables, replication between different major version PostgreSQL and even online major version upgrade. Now we have a way to realize these demands in the form of Logical Replication.
However, because NTT group companies develop various social infrastructures and place a high priority on safety, it's not enough to simply ask to start using a new feature, without clarifying how it works, finding and providing information about various pitfalls, and providing instructions about how to operate it.
With that in mind, we have investigated the architecture of logical replication in detail and tested its various aspects including operation and monitoring.
In this talk, we will share the knowledge we gained during these investigations and various tests.
Especially, we are going to talk about following topics.
architecture and pitfalls
- cases when walsender consumes much more resources than you expected
- cases when response time for publisher increases
- prerequisites for starting up logical replication
- backup and restore under logical replication
- conflict detection and resolution
- safety of using pg_replication_origin_advance()
- what and how of logical replication monitoring
- only monitoring pg_stat_replication may be insufficient
- features that should be added in future
- 2018 April 19 11:50 EDT
- 50 min
- PostgresConf US 2018
- Operations and Administration