[erlang-questions] Frequent Node Disconnects

[Scott Lystig Fritchie]

Jan Vincent <jvliwanag@REDACTED> wrote:

jvl> I find my erlang nodes in a distributed system frequently being
jvl> disconnected under light load. What symptoms should I look out for
jvl> why this happens?

Sorry, you didn't mention if you were using virtual machines or not.
Xen, VMware, and others frequently have problems with guest OS clocks
not keeping anywhere close to the hosts's OS clock, much less "real"
time.

If your Erlang nodes have different values for -kernel net_ticktime,
you're almost guaranteed to have unpredictable disconnection events.

Otherwise, use erlang:system_monitor(), preferably on all nodes, to get
more detail about each specific disconnection.  When on all nodes, you
can compare: did node A and node B agree about the time and/or the
reason for the disconnection?

-Scott