SFRAC 4.1 Oracle clusteraware installation problem

[shashi]

I am trying to install Oracle 10g R2 Clusterware on HP-UX 11.23 with SFRAC 4.1


I have 3 nodes in the cluster.


cayman:/ #cfscluster status

Node : cayman
Cluster Manager : running
CVM state : running
MOUNT POINT SHARED VOLUME DISK GROUP STATUS
/u01 Oracle10g_CodeTree oradg MOUNTED


Node : charger
Cluster Manager : running
CVM state : running
MOUNT POINT SHARED VOLUME DISK GROUP STATUS
/u01 Oracle10g_CodeTree oradg MOUNTED


Node : invicta
Cluster Manager : running
CVM state : running
MOUNT POINT SHARED VOLUME DISK GROUP STATUS
/u01 Oracle10g_CodeTree oradg MOUNTED



cayman:/ #gabconfig -a
GAB Port Memberships
===============================================================
Port a gen 8fb201 membership 012
Port b gen 8fb203 membership 012
Port d gen 8fb206 membership 012
Port f gen 8fb211 membership 012
Port h gen 8fb205 membership 012
Port o gen 8fb207 membership 012
Port v gen 8fb20d membership 012
Port w gen 8fb20f membership 012
cayman:/ #


cayman:/ #lltstat -nv
LLT node information:
Node State Links
* 0 cayman OPEN 2
1 charger OPEN 2
2 invicta OPEN 2


SSH and RSH is configured in all three nodes. SSH/RSH works without password.


Started the Oracle 10g CRS installation on the master node.

First, it didn't identified the nodes which are part of cluster. What will be the cause for this ?. So, i have provided the node names, it private names, and its virtual names through cluster configuration file.

Second, when i have provided the ocr file on top of my cluster file system where i have mounted that filesystem on /u01 on every node, i am getting the error "SEVERE: The location /u01/crs/oracle/product/10gr2/ocr, entered for the Oracle Cluster Registry (OCR) is not shared across all the nodes in the cluster. Specify a shared raw partition or cluster file system file that is visible by the same name on all nodes of the cluster."

Is this indicates some library/kernel module issues with my SFRAC installation ?


But i can able to see the cluster FS mounted file system from all the nodes.

[rsharma1]

1.check for the ownership and permission of the ocr file for oracle user:group
2. ensure that the passwd-less rsh/ssh work for oracle user as well

[Gaurav_S]

Hello,

Is your /etc/hosts file having separate entries for private NICs used in CRS ? also the virtual IPs used by RAC ?

Regarding the /u01 filesystem, what was the activation mode ? did you set the same to shared write ?

Gaurav

[Eric_Gao]

As mentioned above, could be the permission problem, as with the installation of clusterware, CSS will interface with vendor cluster software and make sure it is a real cluster.

can you pls manually run cluster verify utility to see what is the error messages,  one of check is against the cluster file system as below

cluvfy comp cfs -f /u01 –n cayman,charger,invicta -verbose   (as oracle user)