So I'm having a recurring problem where members of my cluster silently drop
off. Their beam.smp process is doing about 1/10 the work of healthy nodes.
This looks like it is related to handoff_concurrency.
After seeing a large number of log messages like:
An outbound handoff of partition
Drew,
That message is normal and is a result of handoff throttling done with the
riak-admin transfer_limit setting. When you say nodes are silently dropping off
what do you mean? Is the beam.smp process shutting down or are they just never
finishing this handoff?
Thanks,
Brian
--
Brian