Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: Oracle 8i on TruCluster
This note applies to Oracle release 8.0, differences with 8i are
noted where appropriate(*)
Introduction
Select (Yes) to install the products on all the nodes that comprise the cluster for the ORACLE8 Parallel Server. Select (No) to install products on this node only. Select (Help) for further assistance. (*) Wording may be slightly different with 8i.
Selecting Yes to this option allows you to run the installer once and it will install Oracle on all nodes in the TruCluster. This can save time, however there are some restrictions. This note discusses these.
Restrictions
+ The installer will prompt you for each node to install the software on:
Enter remote node name on remote node. The remote node name must be the hostname of a node on the cluster. Leave the entry blank and press [RETURN] when you have entered all the nodes on the cluster.
Once you've answered this once for each node, blank out the entry before pressing RETURN.
+ The installer will use the SAME value for ORACLE_HOME on all nodes.
Consequently if you want the following setup:
NODE1 ORACLE_HOME=/oracle1/app/oracle/admin/V806 NODE2 ORACLE_HOME=/oracle2/app/oracle/admin/V806
ie ORACLE_HOMEs are different, you will either have to do separate installs
or setup symbolic links so that /oracle1/app/oracle points to /oracle2/app/oracle on NODE2. After the installation completes the symbolic
links should be removed and the init<SID>.ora and config<SID>.ora updated
accordingly on NODE2.
Our recommendation is if the values for ORACLE_HOME will be different then use a separate install.
+ If the installer is doing multiple installs, the way it works is to
do the installation on the node where the installer is running first.
When
this installation is finished it copies the ORACLE_HOME to the other nodes.
To do this you will need to have your $HOME/.rhosts file correctly setup on ALL nodes in the cluster, so that your installer can copy the files across. The format is:
<remote_machine_name> <username>
eg
ukaa26.uk.oracle.com oracle
Once setup, ensure the remote access is working: From the machine where the installer is going to be run from issue:
% rsh <othernode> ls
If you don't get a directory listing or are prompted for a password then .rhosts is not setup correctly. Please see your UNIX System Administrator
before going any further.
+ As discussed in the previous section, the install is done on subsequent
nodes by COPYing the ORACLE_HOME across. This means that Oracle
does NOT get relinked on the other nodes. After an install completes we
would recommend you perform a manual relink of AT LEAST the Oracle
executable on the other nodes. This can be done as follows:
<login as the Oracle UNIX owner> <shutdown the instance> % cd $ORACLE_HOME/rdbms/lib % make -f ins_rdbms.mk ioracle
The link must complete without errors before you can restart the instance.
It is much better to find out the software does not link now, rather than
X months down the road when (say) installing a patch on a production system.
+ The database created with have the same database name (db_name) as
the SID name supplied for the node where the install is running. For
example
if you are going to have
NODE1 SID=V806A NODE2 SID=V806B
If you run the installer on NODE1 and create a database, the database name given will be V806A.
CONFIGURING TRUCLUSTER ON DIGITAL UNIX
Questions and Answers:
From the boot prompt issue mc_diag and mc_cable. See the Memory Channel User's Guide for more details.
2. Where is the TruCluster Software?
Digital UNIX Associated Products Volume 2 CD.
3. Are there Pre-requisits for loading the Trucluster software?
Yes, along with the memory channel hardware you need Licenses. You need the Trucluster license TCR-UA to be active. But if you are loading all Trucluster subsets you need the following licenses *disabled*: ASE-OA and MCA-UA
Loading the TruCluster Software:
# setld -l /cdrom/TruCluster/kit
3. During Configuration of the TruCluster Configuration Software subset
you are:
An example entry:
IP name IP Address
ukaa25 installation: ukaa25_MC 10.0.0.4 ukaa26 installation: ukaa26_MC 10.0.0.5 The IP address needs only to be visible to the machines. It does not need to be externally visible. c) Prompted for: Will the system be in an ASE [Y}: Answer: 'Y' d) Prompted for: Enter the ASE_ID number [0]: An example reply: ukaa25 id 0 ukaa26 id 0 These values are stored in /etc/rc.config e) Prompted for: Do you want to run the ASE logger on this node? [n]: An example reply: 'N' Installation should now be complete, and you will be prompted to rebuild the kernel.
Configuration:
Run From one node:
% asemgr
You are prompted to:
| Enter a comma-separated list of member systems's interfaces on | the cluster interconnect within ASE 0. An example entry: ukaa25_MC, ukaa26_MC | Enter Members: ukaa25_MC, ukaa26_MC | | Member List: ukaa25_MC, ukaa26_MC | | Is this correct (y/n) [y]: Enter: 'Y' Now enter: % asemgr -dh An example return: | Member Status | | Member: Host Status: Agent Status: | ukaa25_MC UP RUNNING | ukaa26_MC UP RUNNING The interconnects are now configured.
2. Configure a tie-breaker disk.
First create a fail-over service, enter:
% asemgr
m) Managing ASE Services c) Service Configuration a) Add a new service tie_1 rrzc16c Now specify a disk to act as the tie-breaker, enter: % cnxset -d /dev/rrzc16c You need to do this manually on ukaa25 and ukaa26. Confirm with: # cnxshow An example Cluster View: Director: ukaa25_MC Suspended: No Node monitor using tie-breaking disk: /dev/rrzc16c Hostname Cluster I/F CS_ID Incarnation Comm Member Okay ------------------------------------------------------------------- - ukaa25.uk.or ukaa25_MC 0001,0002 0000000000060a83 Yes Yes ukaa26.uk.or ukaa26_MC 0001,0001 0000000000043cef Yes Yes
Installation Validation:
% cnxshow % cnxshow -v This should report no errors. Also use: % clu_ivp This should report no errors. 2. You MAY need to set the follow-up boot parameters at the halt chevron '>>' prompt. Example: shutdown -h >> set boot_reset ON >> set bus_probe_algorithm new *Note: clu_ivp will indicate if this is necessary.
Chi Sim Tang <chisim_at_amoeba.com.sg> wrote in message news:81j5i8$lit$1_at_coco.singnet.com.sg...
> Hi, > > We are trying to install Oracle 8i (v8.1.5.0.0) on a 2-node Compaq > TruCluster running Tru64 Unix 4.0F with a shared channel. We intend to doa
> active-active setup where both nodes are running a Oracle instance and > handling their own databases but would failover via TruCluster if a node > dies so as to achieve high availability. > > For node 1 which is expected to be heavily utilised, we created shared > volumes, ora1_02 (the datafiles) and ora1_03 (the redo logs, control files > and archive) on separate disks. The Oracle software is installed onora1_01
> which is a local volume. > > For node 2 (not expected to have heavy usage), we have ora2_01 which is > local containing the software, and ora2_02 which is a shared volume to > contain all the datafiles, logs, etc. > > The DBAssist program has problems creating custom installation, so we > installed typical and then moved the datafiles and log files manually tothe
> respective volumes. > > We have no idea how to proceed further at this point and are wondering if > anyone has done the same or similar setup and could offer any advice tohelp
> us along. Any gotchas we need to be looking out for will be really > appreciated. :) > > TIA! > > > Chi Sim > > > >Received on Thu Nov 25 1999 - 12:30:09 CST