I'm trying to build up native postgresql 9.2 streaming replication in the Amazon cloud.
According to documentation, I've set up everything correctly – but at the slave I get these logs:
2013-02-04 13:18:01.315 GMT LOG: streaming replication successfully connected to primary
2013-02-04 13:23:07.109 GMT LOG: invalid record length at 0/10000080
2013-02-04 13:23:07.109 GMT FATAL: terminating walreceiver process due to administrator command
2013-02-04 13:23:07.209 GMT LOG: invalid record length at 0/10000080
2013-02-04 13:23:12.214 GMT LOG: invalid record length at 0/10000080
2013-02-04 13:23:17.219 GMT LOG: invalid record length at 0/10000080
2013-02-04 13:23:22.225 GMT LOG: invalid record length at 0/10000080
2013-02-04 13:23:27.230 GMT LOG: invalid record length at 0/10000080
2013-02-04 13:23:32.235 GMT LOG: invalid record length at 0/10000080
2013-02-04 13:23:37.240 GMT LOG: invalid record length at 0/10000080
2013-02-04 13:23:42.245 GMT LOG: invalid record length at 0/10000080
2013-02-04 13:23:47.251 GMT LOG: invalid record length at 0/10000080
This starts, when I insert data into the master. Any ideas what the problem may be?
Best Answer
I've solved the problem. The trick is, not to start the database without a recovery.conf file. Postgresql does something that breaks replication, even if no obvious change was done to the database.