Veritas Cluster Server (VCS) does not detect Oracle resources as online if Oracle binaries are located inside of a Solaris local zone and mounted via NFS from another machine
VCS DBG_2 V-16-50-0 Netlsnr:apo_pu_listener1:monitor:stat call failed for proc file (/proc/28101/object/a.out), errno (5) Netlsnr.sun.C:netlsnr_monitor
During the Oracle agents monitor procedure the agent attempts to stat Oracle binaries to obtain information about the owning user account. This user account is then used for process matching later in the monitor procedure. Without this user information monitoring may fail.
When Oracle binaries are located in a Solaris local zone and NFS mounted, however, users in the global zone on the NFS client (i.e. the machine/zone running VCS) do not have permission to stat files mounted from and NFS server in a Solaris local zone. As such the stat of the Oracle binaries fails and owning user account information is not obtained.
Oracle binaries should be NFS mounted into the local zones zone root from the global zone instead of the local zone. For example, given a zone name of 'testzone' and a mount point of '/opt/oracle' within the local zone, rather than mounting oracle binaries from within the local zone using:
mount server:/filesystem /opt/oracle
The file system should be mounted from the global zone inside of the local zones zone root, i.e.:
mount server:/filesystem /testzone/root/opt/oracle
Once done agents running in the global zone will then have permissions to stat Oracle binaries within the NFS mount.
Note that a final fix for this issue will be included in the 5.1 RP2 Oracle agent.
Storage Foundation HA/Veritas Cluster Server 5.1 RP1
Solaris 10 SPARC/x86
Oracle 10gR2 or later with Oracle installed/running in a Solaris local zone