If I was to believe Metalink Note 368916.1 (see previous post), then the HACMP software must be installed on the server we were installing 9.2.0.8 on. However, having reviewed /etc/inittab and investigating whether cluster services were running, it looked like HACMP wasn’t in use. We checked with one of our Sys Admin colleagues and he confirmed that. I took a trawl through the OUI logfiles for various installations and noticed this in one of the earlier log files.
CRS clusterware not found. Defaulting to vendor clusterware.
Cluster nodes cannot be retreived from the vendor clusterware
(/tmp/OraInstall2005-09-06_01-18-31PM/oui/bin/lsnodes:
cannot initialize cluster interface ).
This system will not be considered as a vendor clusterware.
oracle.ops.mgmt.cluster.ClusterInfoException:
/tmp/OraInstall2005-09-06_01-18-31PM/oui/bin/lsnodes:
cannot initialize cluster interface
at oracle.ops.mgmt.cluster.ClusterInfo.getNodeNames(ClusterInfo.java:269)
[SNIPPED]
at oracle.sysman.oii.oiic.OiicInstaller.main(OiicInstaller.java:628)
This is not a cluster system.
Node selection page will be shown because a cluster install is detected
on the selected home.
Setting the 'ClusterNodes ( CLUSTER_NODES )' property to 'node284,'.
Received the value from the UI page.
Very confusing. So there’s no vendor clusterware available and this is not a cluster system but because of a previous cluster install, Oracle wants to install RAC. When I looked in the Inventory, there was evidence of RAC being installed. (N.B. I’m only showing the 9.2.0.8 home here. The other homes looked similar.)
<HOME NAME="Oracle_9208" LOC="/oem/oracle/product/9.2.0.8" TYPE="O" IDX="14">
<NODE_LIST>
<NODE NAME="node284"/>
</NODE_LIST>
</HOME>
Sadly, this fits in with my software installation experiences so far at this particular site. I’ve only participated in two installations (in 6 months) and both have been less than confidence-inspiring. We’re in the process of moving towards DBAs installing Oracle (rather than Sys Admins), but it looks like there’ll be a bit of an Inventory clean-up required first!
Oh, and when I said in the last blog that the relink to remove RAC had failed, I was wrong. There were just so many warnings about previous exported symbols, ending with a TOC overflow error that things looked bad. After an evening away from it and some research, I realised this was not an unknown problem on AIX.
The 9.2.0.8 installation is fine now and we’ve started patching databases. Having said that, we’re still only working on Dev servers so we have Production servers to come and 10.2.0.4 as well. Looks like there’ll be plenty of scope for more weird installation issues 🙁