Fullsync Susceptible to Timeouts [JIRA: RIAK-2066]

[jcapricebasho]

The keylisting phase of fullsync can take several hours, during which no data is transferred between source and sink.

This makes fullsync susceptible to repeated timeouts in network architectures that are not using TCP keepalive, or do not have TCP keepalive configured correctly across all devices. For example, a default keepalive on nodes of 2 hours, with an intermediate device having a shorter so_timeout.

While this issue is not explicitly a Riak one, not all Riak Enterprise users have the resources to troubleshoot issues like the above, or the access to inspect or alter intermediate devices such as firewalls.