Greg, there's sort of an inconsistency between the developer-oriented streaming replication page and the tutorial.
The streaming replication page says to copy the entire data directory, which would include the WAL logs in the pg_xlog directory. The 5 minute version in the binary replication tutorial skips that directory, the longer one copies it separately, after the backup is complete, so that the slave can perform any committed changes as part of syncing up with the master.
I assume that there is no harm in copying the pg_xlog directory in the 5 minute version, and it was left out as a time-saver. However, if the master fails and the slave has to take over, wouldn't it need the pg_xlog directory? If it hasn't been created, could that cause problems or would the backend know to create it?
--
Mike Nolan
