Im a new hired dba for a banking company. They have two nodes,
the PROD server and the DEV server which are clustered. They also
have a 3rd party apps which is SAP financials. The SAP team
developers where the one who set-up their apps and the ORACLE
184.108.40.206 database on the two nodes. I presumed they clustered it .
As the new dba Im i little bit cunfused about clustered database. As what I understood it is the same as the RAC right? Two instances on
two oracle homes running against 1 shared database? But the set-
up here is quite different. The PROD server it has its own PROD database and PROD instance. and in DEV server has its own DEV
database and instance too, just like stand alone set-up. But the two
are clustered. Is this kind of clustering valid?
My problem now is...im task to install the patch 220.127.116.11 on both
nodes (oracle homes). So i installed first on the DEV node (not
knowing at first that they are clustered...i thot at first they are a
stand alone set-up). I successfully patched the DEV oracle home.
Then I transferred to the PROD node and try to install the patch
again. To my surprise I cant install the patch because is
says "Theres no patch that needs to be installed".
I tried to run sqlplus on DEV node and the version is now 18.104.22.168
while the sqlplus on PROD is still 22.214.171.124.
Can you help me pls how to patch the PROD node? Im afraid to
experiment other options coz its a PRODUCTION db and I might mess
it up. What do u think did i miss here. I read also that if u are to
patch a cluster u have to run the root.sh to all the nodes. So in my
case since i patched first the DEV node, do i need to copy/ftp
the root.sh found on the DEV ORACLE_HOME to the PROD
ORACLE_HOME and run it? Will running the root.sh alone update or
upgrade the sqlplus version on PROD to 126.96.36.199?
As a new DBA you should have been provided with a
Acceptance environment where you'd do this work(which has same setup, although nodes named as DEV and PROD is very dubious).
Anyways apparently you haven't been provided with the same, I presume.
It's a weird scenario anyways you need to investigate if it's the OS which is clustered. I'm guessing that's it's just some TAF kind of setup(from the SAP app dev's point of view) where if the DEV db fails it'll find the DEV schema on the PROD box and vice versa.
But you do need to investigate properly, ask someone for a schematic diagram of what goes where. Beginning right down from the HW and then working your way up to the SW.
I'm a JOLE(JavaOracleLinuxEnthusiast)
--- Everything was meant to be---
So its a valid config to have two separate database on a clustered set-up
I ask the HW engineer and he said that the setup was like what u said. If one node is down then the disk on that node will be transferred to the active node, with some intervention and the ORACLE_HOME of the active node will adopt it, The two nodes are
clustered but not a full-pledged one. Somewhat like half-done HACMP design. The engineer doesnt know oracle
so he cant comment on the clustered db. I ask the SAP dev team and they
said they just run their out-of-the-box db installer and cant comment on
the what to do wth the clustering. They thought also that they have stand alone PROD and DEV set-up too.
But with this clustered ORACLE_HOME, i dont know how to patch the other
home w/o encountering the "Theres no patch need to be applied" error. I
even shutdown the DEV server hoping the PROD will not see that cluster,
but to no avail. Do you have any idea how to turn this cluster off(at least on the oracle side, not the HW side), so that
i can patch the PROD server?
You got it right with your RAC understanding. However, since u have confirmed that there are both a DEV and PRD instance running on each ibm box. And your seniors told u that they are ibm clustered.
Therefore, most likely, its the PRD instance that is going to fail over to DEV if PRD hardware has some issues. IBM HACMP is a hardware clustering failure. Why DEV is there for the fail over is so that the box will not be left idling just waiting for a fail over to take place.
Thanks...maybe is it a failsafe. I dont really understand much the difference about cluster, failsafe, and RAC. I just know they
are synonymous dierivatives of each other. All i know is RAC shared
one database and have more that one instance on separate nodes.
If I have a cluster database with 2 nodes PROD AND DEV,
is there an order to which node is to be patch first? Is there a term
"primary node" and "secondary node"? How can I tell which one
is primary and which one is secondary?
The RAC Patch install procedure says...
1. If u are installing the patch set in a cluster invironment, The
cluster node selection screen will appear.Click each participating
node and install. then run root.sh to all the participating nodes.
Please tell me if my understanding in cluster patch install is right.
1. I can choose 1 and only 1 node to where i can install the patch.
From this node, i will install the patch to all the participating
nodes in a cluster. Once I have chosen that node, I can not go
to other nodes to resume installing the same patch.
It does not tell me which node to install first. So i can assume any
node will do.
...So in may case, since i have patch the DEV node first, then I
will run the patch from DEV node again and click the PROD node
from there? and install the patch? then it will generate a root.sh
on the PROD sever home? even though im installing from another
server (DEV). Will this solve my problem "Theres no patches that
need to be applied"?
How can i detach the a node from the cluster and make it