[zeromq-dev] Heartbeating using TCP keepalives

Bjorn Reese breese at mail1.stofanet.dk
Fri Jan 3 13:45:51 CET 2014

On 01/03/2014 02:03 AM, Alex Grönholm wrote:

> As it stands, the TCP keepalive intervals are given in seconds on the
> vast majority of operating systems.

Furthermore, if the peer is unresponsive then additional probes will be
sent at regular intervals before the TCP keepalive mechanism reports a
timeout. The details are also dependent on the operating system. For
instance, on some BSD derived TCP stacks, 9 probes will be sent 75
seconds apart, so a timeout will not be reported until 11 minutes and
15 seconds have passed.

More information about the zeromq-dev mailing list