wal_sender_timeout should be as long as necessary. Each wal file is 16MB, so it should be *at least* as long as the time needed to transfer 16MB*wal_keep_segments. Take a look at the size of your pg_xlog folder.
On 09/01/2018 09:06 PM, greigwise wrote: > Hello. > > On postgresql 10.5, my pg_basebackup is failing with this error: > > pg_basebackup: could not receive data from WAL stream: server closed the > connection unexpectedly > This probably means the server terminated abnormally > before or while processing the request > > In the postgres log files, I'm seeing: > > 2018-09-02 00:57:32 UTC bkp_user 5b8b278c.11c3f [unknown] LOG: terminating > walsender process due to replication timeout > > I'm running the following command right on the database server itself: > > pg_basebackup -U repl -D /var/tmp/pg_basebackup_20180901 -Ft -z > > It seems to be an intermittent problem.. I've had it fail or succeed about > 50/50. I even bumped up the wal_sender_timeout to 2000. One notable thing > is that I'm running on an ec2 instance on AWS.
The unit for wal_sender_timeout is ms so the above is 2 seconds whereas the default value is 60 seconds(60s in postgresql.conf file).