failover of true cluster....

From: eskhedr <eskhedr_at_gse.ie-eg.com>
Date: Sat, 08 Aug 1998 21:02:15 +0300

hi all,
i have a 2-alpha system (8400 )running DU4.0d with true cluster
production server and oracle parallel server can any body tell me (if
possible) how to configure the production server to be in case of
failover no user have to logout from the server and login again only a
few seconds pause is permitted if this could happen can any body advise
me how it done????

I have explained some scenarios below based on my understanding of TCPS,

using the above mentioned setup. Please can any body tells me if I am
correct in my
understanding or not:
        Scenario 1
        TCPS will incorporate two 8400 servers: S1 and S2. Neither
machine will contain its own memory, instead, both machines will access
memory in the Memory Channel. Oracle Parallel Server will be employed
running two Oracle instances: I1 (controlled by S1's cpu's) and I2
(controlled by S2's cpu's). Lets say, for the sake of an example, that
S1 becomes unavailable. TCPS will pass control of I1 from the now
unavailable S1 to S2, without the users loosing connectivity. There may
be a delay based on the last transactions performed on I1 before the
failure which will have to be "rolled forward" via S2. There are no loss

of processes and the failover is automatic and relatively seamless.

        Scenario 2
        TCPS will incorporate two 8400 servers: S1 and S2. Both machines

will have their own memory as well as sharing some memory on the memory
channel, which will be used to control failover. Oracle Parallel Server
will be employed running two Oracle instances: I1 (controlled by S1's
cpu's) and I2 (controlled by S2's cpu's). Lets say again that S1 becomes

unavailable. Users that were connected to I1 on S1 loose their Oracle
connection. TCPS will start I1 on S2. Users can login again and will be
connected to I1 via S2. Failover is automatic but not seamless.


I hope that I am not asking too many questions at once, but I am seeking

to understand, in detail, the process and options for the failover.

Thanking you in advance


Essam Khedr
Received on Sat Aug 08 1998 - 18:08:30 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:38 NZDT