HACMP started
Before starting Oracle9i RAC instances, the HACMP cluster must be up and running. Check the cluster state with the /usr/sbin/cluster/clstat command.
Hagsuser group
These tests must be performed on all nodes:
Check that the Oracle user is part of the hagsuser group. The name of this group is mandatory, and cannot be changed.
Check, and change if necessary, the permissions on the cldomain executable. This program must be executable by everybody (user, group, other):
# chmod a+x /usr/es/sbin/cluster/utilities/cldomain
Check, and change the group to hagsuser if necessary, for the svcsdsocket.oracle9irac socket file: (assuming that oracle9irac is the name of your cluster, returned by the cldomain command):
{node2:root}/-> chgrp hagsuser /var/ha/soc/grpsvcsdsocket.oracle9irac
Check, and change if necessary, the group permissions for the grpsvcsdsocket.oracle9irac socket:
{node2:root}/-> chmod g+w /var/ha/soc/grpsvcsdsocket.oracle9irac
The HAGS socket file needs to be writable by the Oracle user and the cldomain executable needs to be executable by Oracle. By configuring the group and permissions for the grpsvcsdsocket.oracle9irac file, the instance will be able to communicate with HAGS and the instance will mount.
See Example 3-39 for the valid permissions and group membership for the socket files used by Oracle9i RAC.
Example 3-39 Permissions on the /var/ha/soc directory
{node2:root}/var/ha/soc-> ls -ltotal 0
srw-rw-rw- 1 root haemrm 0 Jun 10 18:18 em.clsrv.oracle9irac
srw-rw---- 1 root haemrm 0 Jun 10 18:18 em.rmsrv.oracle9irac
drwxrwxrwx 2 root system 256 Jun 11 19:20 grpsvcs.clients.oracle9i
racsrw-rw-rw- 1 root hagsuser 0 Jun 10 18:18 grpsvcsdsocket.oracle9i
racdrwxrwx--- 2 root haemrm 256 Jun 10 18:18 haem
drwxr-xr-x 2 root system 256 May 13 17:54 hats
drwxr-xr-x 2 root system 256 Jun 10 18:18 topsvcs
UDP tuning
The udp_sendspace value must always be greater than Oracle9i RAC's db_block_size. Otherwise Oracle9i RAC will not start.
GPFS issues
GPFS requires a quorum of nodes to run. It also requires that both nodes holding the GPFS configuration data be up and running for any configuration changes to be performed.
In case of a system restart, GPFS will be started automatically and the GPFS file systems will be mounted.
To shut down and unmount the GPFS file systems on all nodes, issue:
# mmshutdown -a
To start GPFS and mount the GPFS file systems on all nodes, issue:
# mmstartup -a
Before starting Oracle9i RAC instances, the HACMP cluster must be up and running. Check the cluster state with the /usr/sbin/cluster/clstat command.
Hagsuser group
These tests must be performed on all nodes:
Check that the Oracle user is part of the hagsuser group. The name of this group is mandatory, and cannot be changed.
Check, and change if necessary, the permissions on the cldomain executable. This program must be executable by everybody (user, group, other):
# chmod a+x /usr/es/sbin/cluster/utilities/cldomain
Check, and change the group to hagsuser if necessary, for the svcsdsocket.oracle9irac socket file: (assuming that oracle9irac is the name of your cluster, returned by the cldomain command):
{node2:root}/-> chgrp hagsuser /var/ha/soc/grpsvcsdsocket.oracle9irac
Check, and change if necessary, the group permissions for the grpsvcsdsocket.oracle9irac socket:
{node2:root}/-> chmod g+w /var/ha/soc/grpsvcsdsocket.oracle9irac
The HAGS socket file needs to be writable by the Oracle user and the cldomain executable needs to be executable by Oracle. By configuring the group and permissions for the grpsvcsdsocket.oracle9irac file, the instance will be able to communicate with HAGS and the instance will mount.
See Example 3-39 for the valid permissions and group membership for the socket files used by Oracle9i RAC.
Example 3-39 Permissions on the /var/ha/soc directory
{node2:root}/var/ha/soc-> ls -ltotal 0
srw-rw-rw- 1 root haemrm 0 Jun 10 18:18 em.clsrv.oracle9irac
srw-rw---- 1 root haemrm 0 Jun 10 18:18 em.rmsrv.oracle9irac
drwxrwxrwx 2 root system 256 Jun 11 19:20 grpsvcs.clients.oracle9i
racsrw-rw-rw- 1 root hagsuser 0 Jun 10 18:18 grpsvcsdsocket.oracle9i
racdrwxrwx--- 2 root haemrm 256 Jun 10 18:18 haem
drwxr-xr-x 2 root system 256 May 13 17:54 hats
drwxr-xr-x 2 root system 256 Jun 10 18:18 topsvcs
UDP tuning
The udp_sendspace value must always be greater than Oracle9i RAC's db_block_size. Otherwise Oracle9i RAC will not start.
GPFS issues
GPFS requires a quorum of nodes to run. It also requires that both nodes holding the GPFS configuration data be up and running for any configuration changes to be performed.
In case of a system restart, GPFS will be started automatically and the GPFS file systems will be mounted.
To shut down and unmount the GPFS file systems on all nodes, issue:
# mmshutdown -a
To start GPFS and mount the GPFS file systems on all nodes, issue:
# mmstartup -a
Comments
Post a Comment