PgCon2014ClusterSummit
Clustering and Replication Developers Summit pgCon 2014
Tuesday, May 20th
2PM to 6PM
Preceded by PostgresXC Meeting 9AM to 1PM
Followed by the PostgresXC Pizza Demo at 7pm.
University of Ottawa
Room: University Center Room 205
Sponsored by NTT Open Source
Agenda
1:00PM to 2:00PM
Lunch supplied for anyone coming for the whole day (PostgresXC + Clustering)
Due to a sudden increase in numbers, only the 15 participants staying for the full day (both XC and Clustering) will be given lunch. People attending the clustering summit only should eat their own lunch, and arrive at 2pm.
1:30PM to 2:00PM
Please bring any last-minute agenda items to Josh Berkus at this time.
2:00PM to 2:45PM
Introductions, and status reports from Replication/Clustering Projects:
Status updates (please volunteer below if you can give an update):
- pgPool II: Tatsuo Ishii
- Postgres-XC: Koichi Suzuki
- Built-in Replication: ????
- Postgres-XL: Mason Sharp
- Stado: Mason?
- Bucardo: ???
- LSR & BDR: ???
- Slony: Chris Browne/Jan Wieck
If you are at the summit representing a specific replication or clustering tool, you should prepare a 1-3 minute summary of current progress and issues. If you want to use slides, please provide slides in PDF form to Josh Berkus by Friday, May 16.
2:45PM to 3:15PM
Summary of Clustering API projects.
Summit attendees who have been working on core clustering features should give an update as to progress and current issues. Please present a 5-10 minute summary. Attendees may use their own laptops for slides, or give slides to Josh Berkus.
Topics:
- Logical Replication (not present)
- Exportable Parser/Planner: no activity for year. Some discussion.
- FDW and SQL/MED
- Function scan push-down
- Node partitioning with constraint exclusion
3:15PM to 3:45PM
Break
3:45PM to 4:30PM
Summary of Clustering API projects, continued.
4:30PM to 6:00PM
Discussion of priorities, progress and ideas for core clustering projects and APIs.
Goal of this discussion is to modify the list of core clustering features and get commitments for hackers to work on specific features. Also, to supply discussion items for the following day's Developer Meeting
RSVP List
- Josh Berkus
- Koichi Suzuki
- Tatsuo Ishii (SRA OSS)
- Haruka Takatsuka (SRA OSS)
- Yugo Nagata (SRA OSS)
- Mason Sharp
- Bruce Momjian
- Chris Browne
- David Wheeler
- Moshe Jacobson (Nead Werx)
- Chris Autry (Nead Werx)
- Joe Conway
- Steve Singer
- Jan Wieck
- Jim Mlodgenski
- Vik Fearing
- Marc Linster (EDB)
- Ahsan Hadi (EDB)
- Zahid Iqbal (EDB)
- Etsuro Fujita (NTT OSS Center)
- Amit Kapila
- Krzysztof Nienartowicz (ESA)
- Hitoshi Hemmi (NTT OSS Center)
- Álvaro Hernández
- Motoyuki Kawaba
- Haribabu Kommi
- Yurie Enomoto
RSVPS to the Cluster Summit are Closed
Meeting Notes
Introductions
We did not capture names of attendees. Simon Riggs and Andres Freund were not able to make the meeting.
Projects
- pgPool II: Yugo Nagata
- 3.4 improvements, adds PG exception and memory manager
- slides
- pgPool II: Ahsan Hadi
- performance testing
- scaling improvements for 1-5 replicas and 75% read workload
- Postgres-XC: Koichi Suzuki
- release 1.2 items
- more openness for volunteers, committers, roadmap
- Stado: Mason Sharp
- nothing big to report
- move from Launchpad to Sourceforge
- Postgres-XL: Mason Sharp
- optimizations over Postgres XC
- focus on stability, performance, and openness
- Slony: Chris Browne/Jan Wieck
- Slony top added
- better tooling
- add logical replication support to avoid trigger overhead, have prototype
- LSR & BDR: Mark Sloan
- Intel funding
- still work for 9.5, e.g. DDL
Clustering API Projects
- Passing Query parse trees to other nodes
- no one volunteered to work on it
NTT is working on some features for FDWs:
- FDW
- add syntax to allow FDW partitions/inheritance
- this was intended for 9.4, but got pushed back
- support constraint_exclusion across shards
- add syntax to allow FDW partitions/inheritance
- FDW pushdowns
- JOINS
- Aggregates
- Scans
- Contributing XC/XL code to mainstream PostgreSQL.
New Projects
The meeting discussed a number of possible new projects. The below lack anyone qualified to work on them:
- Async FDW execution
- Parallel Query?
- GTM
- Hooks for the GTM?
- Two-node sync?
- WAL barrier for shards
- Problem with transaction commit order visibility differences between master/slave
- Partition primary keys/foreign keys
- either using the partition index, or not
- other multi-table constraints
There was some discussion of the new BDR features, hampered by the absence of the BDR team. This included:
- Adding more simple BDR formats to core/contrib
- Adding a core/contrib LSR apply plugin, maybe based on Slony copy.
We discused the need for multiple host connections for HA configurations. For example, if you have a pgpool and a failover pgpool, you'd like to be able to connect to both. Jim Mlodjenski committed to working on a multihost connection change for libpq, which would look like:
database=postgres host=10.0.0.1,10.0.0.2,10.0.0.3 port=5432
PostgresXC Pizza Demo
Koichi Suzuki demonstrated setting up and managing PostgresXC with the new pgxc_ctl. This program, despite its name, is a full command shell for a PostgresXC cluster, supporting node changes, sharding commands, and status checks.