Thanks to all that responded. My original query was:
>I am setting up several services on my 2-member TruCluster
>1.6 ASE server but I am very limited in the number of IP
>addresses allocated for my use. Will I run into a problem
>if I give the services unique names, but alias all
>of those names to the same IP address in the /etc/hosts file?
Everybody was in agreement that this could not be done if I
expect the services to run on different machines. The most
practical alternative for me is to combine all of my disk
services into one service so that they will always run on
the same machine.
Here are some of the responses:
-----
Thomas Strandenęs wrote:
a lot of trouble. if two services try to live on separate nodes, they will
get an IP address confict, and while the machines fight for the right to
service the IP number in question, your services will fail miserably.
what you want to do instead, is to run several programs *in the same
service*. hence, you'll write scripts to start and stop several programs in
the same service, and they'll share the same IP and live happily ever after.
-----
Davis, Alan wrote:
Yes, this will be a problem. Each service /must/ have
a unique IP address in order to failover properly.
The system will not allow you to define multiple services
with the same IP.
-----
Andrija Pantovic wrote:
It is enough to assign IP address to only one service (the "ifconfig alias"
command will be performed once), there is no reason to assign same IP
address to
all services.
Ofcourse, this means that all services are on the same member and failover
is
performed for all of them in the same manner.
-----
Van Bever, Pascal wrote:
Don't do this!
Indeed, as there is no hierarchy/dependency between ASE services, you cannot
garantee that service x will always run on the same system as service y.
So you need one IP address per service on each network you want to be able
to reach your service.
If you are really stuck with the 1 IP address, you need to create 1 big
service containing all disks/applications you have.
Or go for the dirty way: don't associate any address with the service (can't
do this for a nfs service!) and instruct your users to try both native IP
addresses of the 2 systems in turn.
This really depend on the behaviour of your application.
e.g. you have a client/server application listening on port *:nnn the server
daemon is started by the service start script let say on host b.
client try to connect host_a:nnnn --> failure than try host_b:nnn -->
Received on Wed Feb 09 2000 - 11:48:15 NZDT