Refer to the HP MC/ServiceGuard Manual for more detailed information
/etc/cmcluster/cmclconf.ascii: Defines members of the cluster, lan interfaces, shared disks, and all other global resources for the cluster. To apply modifications, halt all packages, halt the cluster, and run cmapplyconf -v -C /etc/cmcluster/cmclconf.ascii.
/etc/cmcluster/packagedir/package.conf: defines package-specific parameters. Includes SUBNET directive, which should be set to the network where the "floating" package IP address resides. Do not edit this file while the package is running. To apply changes, halt the cluster and run cmapplyconf -v -P /etc/cmcluster/packagedir/package.conf. NOTE: The package.conf files must be identical on all servers in the cluster.
/etc/cmcluster/packagedir/package.cntl: control script for the package. Do not edit this file while the package is running. To make changes, halt the package, edit the file, and restart the package. NOTE: The package.cntl files must be identical on all servers in the cluster.
Commands: (note that all commands should be run with -v option for verbose
cmviewcl -v : shows status of the cluster, including lan interfaces, status of all nodes, current location of packages, critical subnet monitoring, and whether failover is enabled for each package.
cmapplyconf -v -[C|P] filename: used to create/modify Serviceguard binary files. Required after changes to cmclconf.ascii (-C option) and package.conf files (-P option). Should not be run while cluster is active.
cmhaltpkg -v package_name : halts package.
cmrunpkg -v package_name : runs package on the node where command is executed. Use cmrunpkg -n node_name -v package_name to specify a different node.
cmmodpkg -v -e package_name : enables failover for the specified package. You can see if failover is enabled using the cmviewcl -v command.
cmhaltcl -v : halts the cluster. Use cmhaltpkg first to halt each package.
cmruncl -v : starts the cluster and all packages
Working with shared disks
In some situations, you will need to access shared cluster disks without the associated package running. This is common with Oracle database setups, where stopping Oracle will trigger a package shutdown and leave the package disk(s) inaccessible. Follow these steps:
· Identify the shared volume group for the package - grep VG /etc/cmcluster/package_name/package_name.cntl.
· Identify the lvols for each shared VG and their mount points
· Halt the package
· For each shared VG:
· vgchange -c n /dev/vgXX
· vgchange -a y /dev/vgXX
· Mount each lvol to appropriate mount point
Once the work is done, the VGs must be returned to cluster mode:
· umount each lvol from the shared VG
· vgchange -a n /dev/vgXX
· vgchange -c y /dev/vgXX
· cmrunpkg -v package_name
If the VGs are reluctant to enter cluster mode, you can also halt the cluster, run cmapplyconf -v -C /etc/cmcluster/cmclconf.ascii, and restart the cluster. This will reset the clustering flag on all VGs.