Re: Registering a virtual IP address from a C program

Hi Jeff,

I'm not running the application on all of the cluster nodes at the
same
time because it is a database application and the customer has
decided that they do not want to run it in a cluster configuration due
to the distance between data centers.
<<<<<<<<<<<<<<<<<<<<<

I think the point that Frank was making is that latency concern is based on
bull*** (or at least the configurations, DLM options, and hardware
performance of yesteryear)

Admittedly, if you're running Rdb and have talked to anyone in their
engineering teams over the last 10 years then they would have eagerly told
you that "clusters are ***; use row-cache and single nodes". Advice that
has certainly neutralized one of VMS's main advantages and seen-off many a
VMS customer :-(

Anyway, fear not! It is Row-Cache that is the steaming pile o' pooh that no
other DBMS has implemented (or ever will implement). Look at the DLM
enhancements, look at new cluster interconnects, and jump into the
beautiful, cost-effective, world of disaster-tolerant clusters!

OTOH, if the distance between the nodes is architecturally prohibitive
(something that is getting harder to do) or the DB design is such that it
confounds even the biggest, most modern configurations, then sorry I don't
know how to do what you ask in a 3GL and forget I said anything :-)

Regards Richard Maher

"Jeff55" <jeff3140@xxxxxxxxx> wrote in message
news:9512362c-76b3-4033-b446-1d42f73bebd2@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
On May 29, 6:09 pm, sapienzaf <sapie...@xxxxxxxxxx> wrote:
On May 29, 2:46 pm, Jeff55 <jeff3...@xxxxxxxxx> wrote:

Other suggestions?

All of that failover that you're trying to implement has already been
done for you. Both in the cluster IP support and FailSafe IP.

Why have you chosen not to run the application on all cluster nodes at
the same time? I understand the use of hot standby in other
situations, but you have an opportunity with clusters to make your
life much easier.

I'm not running the application on all of the cluster nodes at the
same
time because it is a database application and the customer has
decided that they do not want to run it in a cluster configuration due
to the distance between data centers. Instead they will use a
replication
process that tolerates high latency connections with less effect on
the
active users. In any case, there are two VMS boxes each serving
a local population and serving as a backup to the other.

The problem with FailsoftIP and Cluster IP support is that they don't
include us, the application, in the failover process. If we detect
some
kind of internal problem we may decide that its time to fail over to
the
backup system. Everything might look fine to the o/s monitors though.

I can script something using ipconfig where I create/start/stop an
interface
on the "active" node to host the assigned "virtual IP" address we use
to
follow the application. However this requires privileges so I'd
rather to it in C
because then I don't have to worry about someone tampering with my
script
that I'm going to invoke with privileges. If the source to ifconfig
was available,
that would be perfect.

OpenVMS Service Control seems interesting. I'll take a look at the
source
and see if I can glean any clues.


.