On 11/7/15 12:47 PM, Arnt Gulbrandsen wrote:
> Miles Fidelman writes:
>> Not for nothing, but if you're coordinating distributed servers, your
>> system design is WAY too closely coupled if boot time effects anything.
>
> Boot time is just a kind of downtime. If downtime lasts too long, the
> health checks declare nodes bad and expensive recovery starts.
>

Actually, in my experience, the problem is when downtime is too short,
and worse, when things flap. With a hot failover process, once one has
failed over, one wants to wait a long time before failing back to the
primary system. Otherwise there's a serious risk of flapping back and
forth - which is REALLY expensive. One wants lots of hysterisis in the
system.

Miles Fidelman

--
In theory, there is no difference between theory and practice.
In practice, there is. .... Yogi Berra