============================
* Support for Telegraf statsd
* Check cluster state concurrently from all endpoints
* Don't collect replication log metrics or issue warnings about it until we
have actually seen a master. Previously pglookout could warn about
replication lag while PostgreSQL was performing archive recovery.
* New `primary_conninfo_template` configuration option to generate a
primary_conninfo for recovery.conf when we start following a (new) master.
Previously pglookout tried to read an existing primary_conninfo and
replace the host part of it with the new master's address, but this may
not always work, for example in case there's no primary_conninfo at all.
* Automatically promote one of the standby nodes if pglookout has never seen
a master for N seconds
* Failover now selects the node with highest identifier if multiple nodes
have the same replication position
* `remote_conns` object's keys no longer have functional meaning. Remote
addresses are only read from the values of the object.
* Show version number in pglookout --version
* Emit a warning when running under systemd without python-systemd installed
* Fix logging to syslog and use the same format when logging to journald
* Drop support for Python 2.6
* Various minor bug fixes, refactoring and improved test cases