[Spread-users] switch failover and spread token lose

[Weinsberg Yaron]

Hi Yair,

Our configuration consists of 4 machines on the same segment.
Each machine hosts 2 Ethernet cards that are connected to different
switches.
The switches are connected together in order to provide HA.

The problem is that when a machine crashes/get partitioned the switches are
aware of that
only after 2 seconds (I guess that the switches update their ARP tables
every 2 sec). The thing is that
spread will come to the conclusion that the network is partitioned without
giving the chance for the switches to
revive (by updating everybody's ARP tables).

What can be done except making spread's token timeout above this threshold ?
(which is not deterministic...)

    best,
        yaron.