RAC questions
- From: johan Eriksson <johan.eriksson@xxxxxxxxxxxx>
- To: oracle-l@xxxxxxxxxxxxx
- Date: Fri, 20 Jan 2006 13:18:47 +0100
Hi all,
Our setup is oracle 10.2 on Red hat linux 3 and some connections came
from an JBoss application server.
We are testing our RAC system and it has resulted in some questions and
problems. So if anyone could explain the behaviour, point me to relevant
documentation or/and recommend the best book(s) about RAC I would be
happy man again.
The first problem we have are:
They did bring down the interface that interconnect was using on one of
the nodes while running queries on both the nodes. On node1 the query
was executed as it should and on node2 it hanged (was on this node they
disabled the network interface).
When checking with ifconfig the VIP adress from node2 had been
transferred to node1 (this is how it should work if I understand it
correctly) and on node2 the VIP was gone.
But when we bring back eth1 on node2 again nothing happens, VIP is still
on node1, we have tried to restart everything but nothing happens, node2
it still unavailible...
Should this work automatically or is it manual work that is the intended
solution?
What is the default behaviour if the interconnect goes down?
Another problem we noticed:
During some load we did shutdown one of the nodes and most of the
session went over to the remaining node (some gave errors). But when we
brought back the second node it remained idle, everything was still
executing on the same node. Should oracle spread the load? If so, why
didn't it do it? What actions/changes in configuration do I need to
take?
A question about load balancing:
In my connect string I say (LOAD BALANCE = yes), but what type of load
balancing will i get with this? Is it that it only chooses which
listener it should use randomly? If I connect without this, will my
session still load balanced? For example if my settings in v$session
show FAILOVER_TYPE=NONE, FAILOVER_METHOD=NONE instead of SELECT and
BASIC, what result will this give?
And last a question about interconnect:
What could happen if one node looses the interconnect? Is it possible
that the individual nodes continues to work on the datafiles and somehow
corrupts the data?
Will each part of the RAC sees themself as working (and only the working
part) ?
As you can see, alot of questions, so I would really appriciate
recommendation on what you consider as the best written source about
RAC.
/Johan
--
//www.freelists.org/webpage/oracle-l
Other related posts:
