Enabling round-robin failover with failback

[emollusion]

What section to enhance

failover
Round Robin Load Balancing

Description of the problem

It would be nice to clarify how to use failover with failback to enable a round robin loadbalancing with failover.
Reading the sections separately gives the impression that this is not done, so a guide on how to do this since it is possible would be a nice addition.

Proposed solution

Addding a failover for each destination in the loadbalancing, including the next server(s) together with failback enables loadbalancing between multiple servers. In the below example two destinations are used, so when one of them goes offline all traffic goes to the other, and when both are online the traffic is balanced as normal.

This should be possible to do with any amount of destinations I have however only tested it with two so far.

Config Example for Destination

destination d_syslog_udp {
  channel {
    channel {
      filter {
        "0" == "$(% ${R_MSEC} 2)"
      };
      destination {
        network(
          "first-server.com"
          port(514)
          transport(udp)
          spoof-source(yes)
          spoof-source-max-msglen(1024)
          failover(
            servers("second-server.com")
            failback(
              successful-probes-required(3)
              tcp-probe-interval(60)
            )
          )
        );
      };
      flags(final);
    };

    channel {
      filter {
        "1" == "$(% ${R_MSEC} 2)"
      };
      destination {
        network(
          "second-server.com"
          port(514)
          transport(udp)
          spoof-source(yes)
          spoof-source-max-msglen(1024)
          failover(
            servers("first-server.com")
            failback(
              successful-probes-required(3)
              tcp-probe-interval(60)
            )
          )
        );
      };
      flags(final);
    };

  };
};

[HofiOne]

@emollusion

"It would be nice to clarify how to use failover with failback to enable a round robin loadbalancing with failover."

Even the earlier version (before the last edit) of the documentation stated the following

round-robin mode: If failback() is not set, syslog-ng OSE does not attempt to return to the primary server even if it becomes available. In case the failover server fails, syslog-ng OSE attempts to connect the next failover server in the list in round-robin fashion.

...

failback mode: If failback() is set, syslog-ng OSE attempts to return to the primary server.

So, to get the round-robin mode work, you simply should not use the failback() option at all.
If you provide the failback() option, it will not be a simple round-robin method anymore.