Health Checking.

• New pool member parameter "monitor_ip" allows to specify an IP address different from the member's IP to issue the health checks against, this will allow the users to configure health endpoints on servers that are different from the servers providing the service itself.
• New HTTP monitor parameter "expected_codes" allows to match on an array of HTTP response codes, previously the health check required a healthy response to contain 200 code.
• New monitor type "forced" allows to force all pool members to be either UP or DOWN effectively disabling the health checking.
• Increased the max health check timeout from 5 to 10 seconds to allow to monitor endpoints that may be taking an extended time to respond.

DNS.

• All SOA parameters(e.g. MNAME, RNAME, REFRESH, RETRY etc.) are now configurable.
• Added support for "auto" SOA serial that will set the serial to seconds since UNIX epoch of the last state update.
• Increased the max "max_addr_returned" from 32 to 1024, automatic fall-back to TCP for large responses.

Administration.

• Reworked the startup scripts, all startup operations now require a valid LB configuration to be present to safeguard against configuration errors.
• Improved health status logging messages.
• Reduced the max allowed member weight from 99 to 10 to limit potential size of the generated state tables.
• polaris-memcache-control now takes a mandatory hostname_or_ip as the first positional argument

Internal.

• DNS resolver thread will now pull state updates from the shared memory asynchronous to its resolution loop. Previously the state pull process was done in the same thread performing DNS responses, this resulted in a greater variance in response times and limited configuration size scalability.
• Health tracker will now push state updates into the shared memory asynchronous to its scheduling loop. Previously the state push process was done in the same thread performing the health checks scheduling, this could result in unpredictable delays in the scheduling for larger configurations.
• Improved topology lookup logic, it will now work with overlapping subnets returning the most specific(longest prefix length) match.
• Better shutdown procedure, "stop" command is now supported on the control socket.
• Increased RECV_BUFFER_SIZE from 1K to 8K allowing to read incoming data faster.