Are NoSQL databases more reliable?
summary
Claims a Postgres setup can get NoSQL-style majority-commit reliability by continuously reconfiguring quorum groups instead of simple standby promotion. Built a prototype called Catch (Go) that uses PG rewind, HashiCorp memberlist for discovery and a Paxos-lease-like protocol to create and revoke “epics” (acceptors). Tests run multiple Postgres instances on one host with asynchronous replicas created; synchronous reconfiguration is planned but not yet implemented. Design targets a three-member quorum plus spares to survive extra failures.
AI-written summary · about
chapters (10)
- 0:00Intro
- 0:58What is reliability? NoSQL databases commit updates to a majority within a group. Latest copy can be read even when an instance is down.
- 2:38Is synchronous replication reliable? Traditional synchronous replication breaks the mirror. Only one instance receives updates after a standby is promoted.
- 3:56How does Ketch provide reliability? All updates are commited to a majority within a quorum group.
- 9:25What is Ketch?
- 11:25Running Ketch Servers
- 11:58Connecting to a Database under Ketch
- 12:34Querying Ketch Resources Ketch has multiple resources which can be queried from the CLI. List servers the local instance is connected to
- 15:45Is Ketch ready to use?
- 16:42Other Roadmap Items List of future features in no particular order: • Integration with SkyDNS for service discovery • Operation to force a failover of a single database
related talks
12mNewSQL Leveraging PostgreSQL
26mNewSQL: concepts, realities and when you should embrace it
28mQuorum-Based Consistency for Cluster Changes with CloudNativePG Operator
29mModern Database Architecture: From SQL to NoSQL & Distributed SQL
57mReliability features in Postgres Pro
48mThe NoSQL Way in PostgreSQL
31mSQL to NoSQL to NewSQL and the rise of polyglot persistence Paul Dix
55mHigh Availability with PostgreSQL and Pacemaker