## Bug Fix

### Update minimum_master_nodes on Elasticsearch node removal

This fixes an issue when removing nodes from the cluster would not
adapt the minimum node count required for the Elasticsearch cluster
managed by Chef Backend, leaving the cluster in an unrecoverable
state.

## Enhancements

### Include disk status in health checks, and status output

When the available disk space for the log and data directories used
by Chef Backend falls below a user-configurable value (default:
250MB). By default disk space failures are considered non-fatal for
the purposes of leaderl health checks. That is, low disk space will
not force a fail over. To make disk space failures fatal, set the
following configuration option:

```
leaderl.health_check.fatal_system_checks true
```

The disk-space-related health status can be queried using:

```
chef-backend-ctl status disk
```

which is exposed via the status endpoint.

### Improve robustness regarding EPMD

The Erlang Port Mapper Daemon (EMPD) is part of the Erlang runtime
system. Typically, this service is started automatically by the
first Erlang process to start on a machine. By promoting EPMD to a
service under supervision the service will now appear in
`chef-backend-ctl status` and will be restarted on failure.

  Further, leaderl will now periodically attempt to re-register itself
  with EPMD if EPMD has failed.

## Security Updates

Bump zlib to 1.2.11.