[Spread-users] Question About View ID / group_id

[Ryan Caudy]

Your above guess is incorrect, Nilo.

The group ID consists of a memb_id, and an index.

The memb_id reflects the IP address (proc_id) of the configuration
leader at the time it is selected, as well as a UTC timestamp (with
some fudging to avoid repeats if IDs are generated rapidly) taken at
that host.  The memb_id component of the group_id is changed whenever
a heavyweight membership change affects the group.

The index is incremented with each lightweight membership change.

Hence, in a configuration without IP address or clock changes, there's
no danger of a repeat in the same or different partitions.

Cheers,
Ryan

On 8/30/05, Nilo Rivera <nrivera at cs.jhu.edu> wrote:
> Hi,
> 
> I wonder if someone could explain me how the group_id actually works, and
> how unique is it.  If servers start in a partitioned network, my
> guess is that a common group could have the same or different group_id in
> disconnected sites of the partition, and for each membership change,
> the new group_id equals the largest in the new membership plus one.  I
> noticed is 12 bytes long, so I wonder what additional information is it
> really stored there (or if my guess above is incorrect).
> 
> Thanks,
> Nilo
> 
> 
> 
> 
> _______________________________________________
> Spread-users mailing list
> Spread-users at lists.spread.org
> http://lists.spread.org/mailman/listinfo/spread-users
>