vxfenswap - Assist online replacement of coordination points.
vxfenswap -g diskgroup [ -a autoconfirm ] [-n | -p protocol]
vxfenswap [ -a autoconfirm ] [-n | -p protocol]
vxfenswap -a cancel [-n | -p protocol]
The vxfenswap utility changes the coordination points that are used for arbitration in the event of a network partition. This change happens on a running cluster.
This utility uses /etc/vxfenmode.test file to get the new fencing mode and disk policy from the user (so as to pick DMP (Dynamic Multipathing) paths for the coordinator disks). The format of /etc/vxfenmode.test file and /etc/vxfenmode file is the same. You must create/etc/vxfenmode.test file before running the vxfenswap utility. The utility usesvxfenconfig command to change the coordinator points and or the fencing mode in the vxfen driver.
If the /etc/vxfenmode.test file contains vxfen_mode value set to scsi3, then the online coordination point migration is done in scsi3 mode. If the /etc/vxfenmode.testfile contains vxfen_mode value set to customized, then the online coordination point migration is done in customized mode.
For online migration of coordination point in scsi3 mode, you need to create /etc/vxfenmode.test file only on the node on which vxfenswap utility is executed. The utility uses the file to read the scsi3_disk_policy, and diskgroup specified with -g option to create the new list of coordinator disks.
For online coordination point migration in customized mode, you need to create /etc/vxfenmode.test file on all nodes of the cluster. The utility uses coordination points (servers or disks) specified in the file to create a new list of coordination points. If the new set of coordination points contains coordination points servers, then prior to using the utility, the coordination point servers should be properly set up to serve as coordination points for the VCS cluster.
See vxfenmode(4M) to understand how to specify the servers and diskgroup that are used as coordination points . Also, see "Preparing CP servers manually for use by VCS cluster" section of the Cluster Server Configuration and Upgrade Guide for details on how to set up coordination point servers to serve the VCS cluster.
Before you run the vxfenswap utility, if the /etc/vxfenmode.test file does not exist on a local node, then vxfenswap utility uses information available in the /etc/vxfenmodefile. This ensures consistency in the behaviour of vxfenswap utility with older releases.
The online migration process fails if any of the cluster nodes is not able to register with a majority of the coordination points.
The utility picks up the nodes of the cluster from the/etc/llthosts file and also picks those nodes that have fencing configured. The nodes can also be specified in the /etc/vxfen_nodelist file. The utility makes use of ssh and scp protocols to communicate with nodes. Ensure that communication is set up properly before using the utility. Instead of using ssh and scp protocols, rsh and rcp protocols can be used to communicate with nodes by specifying the -n option.
In some cluster environments, administrators may not want to run the utility using SSH or RSH in the cluster environment. Veritas has enabled hacli support in the vxfenswap utility for such environments. Using the -p command line option, you can specify the protocol for communication with other nodes of the cluster.
By default, hacli is disabled in the cluster. Hence, before you run vxfenswap with hacli, change user level to a root user level by running the haclus -modify HacliUserLevel COMMANDROOT command. After you complete operations with the utility, change the user level back to the disabled state by running the haclus -modify HacliUserLevel NONE command.
See hacli(1M) to understand the usage of hacli commands.
Warning: After the utility is executed, the whole cluster may panic if any node leaves the cluster membership before the utility returns successfully.
-g diskgroup Use diskgroup specified with -g option. Applicable only for online migration to/ within scsi3 mode. -n Use non-secure communication. By default, the vxfenswap utility makes use of ssh and scp to communicate with nodes. When this option is specified, this utility will use rsh and rcp for communication. -a autoconfirm Automatically commit the replacement operation without prompting for confirmation. If this option is not specified, the vxfenswap utility prompts for confirmation before changing the coordination points in the vxfen driver. -a cancel Rollbacks any previously incomplete online coordination point replacement operation. -h Displays the usage for the vxfenswap command. -p protocol Use protocol to specify the communication protocol. The supported values are SSH, RSH, and hacli. This option cannot be used with -n.
For online migration from customized to scsi3 mode with disk group vxfencoorddg:
1. Copy sample vxfenmode file for scsi3 (e.g. /etc/vxfen.d/vxfenmode_scsi3_dmp) to /etc/vxfenmode.test on local node. 2. Run vxfenswap as follows:
vxfenswap -g vxfencoordg
For online migration from scsi3 to customized mode: 1. Use the sample vxfenmode file for cps (/etc/vxfen.d/vxfenmode_cps) to create /etc/vxfenmode.test on all cluster nodes. 2. Run vxfenswap as follows:
For swapping coordinator disks with disk group vxfencoorddg in scsi3 mode: 1. Copy /etc/vxfenmode to /etc/vxfenmode.test on local node. 2. Run vxfenswap as follows:
vxfenswap -g vxfencoordg
For swapping coordination points in customized mode: 1. Use the sample vxfenmode file for cps (/etc/vxfen.d/vxfenmode_cps) and edit it to contain the new set of coordination points. This has to be done on all the cluster nodes. 2. Run vxfenswap as follows:
vxfenadm(1M), vxfenconfig(1M), vxfenmode(4M), hacli(1M)