Heikki,
Tested this on yesterday's snapshot. Worked great.
Test:
4 Ubuntu 10.04 LTS Cloud Servers (GoGrid)
Configuration:
Compiled 9.3(12-12-12)with: pg_stat_statements, citext, ISN, btree_gist, pl/perl
Setup Test:
Master-Master
Replicated to: master-replica using pg_basebackup -x.No archiving.
Master-Replicareplicated to Replica-Replica1 and Replica-Replica2using pg_basebackup -x
All came up on first try, with no issues. Ran customized pgbench (with
waits); lag time to cascading replicas was < 1 second.
Failover Test:
1. started customized pgbench on master-master.
2. shut down master-master (-fast)
3. promoted master-replica to new master
4. restarted custom pgbench, at master-replica
Result:
Replication to replica-replica1,2 working fine, no interruptions in
existing connections to replica-replicas.
Now I wanna test a chain of cascading replicas ... how far can we chain
these?
--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com