On 7/20/06, Alex Collins <[EMAIL PROTECTED]> wrote:
I have run up a test based on the network diagram on page 193 of the
SRSS 3.1 Admin guide.
Is there any reason why I can't just run like this so that either server
will fail over to the other.

If you set these machines up as a host group then, on the average,
half of the Sun Rays at each site will connect to the machine at the
other site.  If you have enough bandwidth between the sites for this
to be acceptable (you say you have plenty of it) then there is one
remaining technical requirement and a couple of operational issues
that you need to consider.

The technical requirement is that in order to function as a host group
the machines need to be able to exchange group membership
discovery traffic.  This is usually sent as multicast traffic but you can
choose to use broadcasts instead of you wish.  Officially the
machines are supposed to have a direct connection to at least one
subnet in common so that broadcast can be used if multicast turns
out to be unreliable but if you're confident that multicast propagates
dependably on your network you might decide to risk ignoring that
requirement.  The archives of this list contain detailed discussion of
this issue.  For starters, run 'utgstatus' on the machines as they
exist today.  That will tell you whether they can already "see" each
other for the purposes of group membership.

The first operational issue is that the Sun Ray Data Store, which
provides the configuration repository for all members of a group, has
a single-master architecture.  One machine is more important than the
others, and if that machine is unavailable then no Data Store updates
can be performed.  (The other machines carry read-only copies of the
DS content: all updates are applied at the master, or "DS primary", and
then fed back to the DS secondaries.)  This isn't usually a big deal but
it will increase your operational risk a little.  You should satisfy yourself
that the risk will be more than offset by the load-sharing and availability
gains before you take the plunge.

The second is that if one machine goes down then, provided that DHCP
is still available, all of the Sun Rays will get a session on the remaining
machine.  Is one machine able to support the full Sun Ray population?

If you're at all unsure about the inter-site bandwidth being adequate
then before you do anything with a host group you could run an
experiment by adjusting DHCP so that it sends all of the Sun Rays at
one site to the server on the other site.  Do this in both directions for
a worst-case scenario.  A few hours of running in that configuration
will confirm whether your bandwidth really is sufficient.  Doing this in
one direction only will place all of the Sun Ray sessions on one
machine, which will tell you whether one machine can handle the
full load.

OttoM.
__
ottomeister

Disclaimer: These are my opinions.  I do not speak for my employer.
_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users

Reply via email to