When using PRIMECLUSTER in a VMware environment, clustering (virtual machine function) between guest OSes on multiple ESX hosts and clustering (virtual machine function) between guest OSes on a single ESX host are available.
When an error occurs on an ESX host or a guest OS within a VMware environment, applications on that guest OS will no longer work. In clustering between guest OSes on multiple ESX hosts, with PRIMECLUSTER applied to guest OSes, when an error occurs, applications will failover from the active guest OS to a standby guest OS in the event of a failure, which creates a highly reliable guest OS environment.
In the cluster environment between guest OSes on multiple ESXi hosts, use the following forcible stop function.
Use SCSI-3 Persistent Reservation as the exclusive control function to panic and stop the error guest OS. By this operation, the operation can be switched. To use this function, Shared RDM (Raw Device Mapping) disk supporting SCSI-3 Persistent Reservation is required.
Note
A forcible stop with the I/O fencing function is disabled in the following environments.
Environment between guest OSes on a single ESXi host
Environment in which the cluster application is configured with 3 or more nodes
Environment in which multiple cluster applications that use a shared disk exist
Information
In the cluster system where the shared disk is connected between guest OSes on multiple ESXi hosts, by setting the SA_icmp shutdown agent, response from the guest OSes is checked on the network paths (administrative LAN/interconnect). The application will be switched when no response is confirmed from the guest OSes. In this case, if the error guest OS does not stop completely (when the OS is hanging, for example), both guest OSes may access the shared disk at the same time. By using SCSI-3 Persistent Reservation, the I/O fencing function prevents both guest OSes from accessing the shared disk at the same time.
Figure G.1 Cluster Systems in a VMware Environment
See
For details on VMware, see the documentation for VMware.
Note
In a VMware environment, the cluster application using shared disks can be configured up to two nodes.
If the guest OS fails in clustering between guest OSes on a single ESX host, the node will become the LEFTCLUSTER state. For how to recover from the LEFTCLUSTER state, see "6.2 Recovering from LEFTCLUSTER" in "PRIMECLUSTER Cluster Foundation (CF) Configuration and Administration Guide." For the following operations, see "7.2 Operating the PRIMECLUSTER System."
When the hvswitch -f command is executed to start or switch cluster applications, the following message is output and starting or switching of the cluster applications may fail.
ERROR: Forcibly switch request denied, unable to kill node <SysNode name>In this case, stop the node "SysNode name" and then execute the hvswitch -f command.
Since cluster systems between guest OSes on multiple ESX hosts in a VMware environment check guest OS statuses via network paths (administrative LAN or interconnect) before effecting a failover, it may happen that guest OSes on which an error occurred are not completely stopped (e.g. when the OS is hanging). Therefore, when using shared disks, you should make sure to set up I/O fencing.
I/O fencing must be set up at the earlier stage of configuring the cluster application.
When using I/O fencing, the shared disk device should be managed by GDS.
To use a switchover file system on the shared disk, you cannot specify the device of the file system to be mounted by defining names such as the label name or the udev functionality in the /etc/fstab.pcl file. Use the device name which can be specified by the entry beginning with /dev/sfdsk.
If a failure in the cluster interconnect occurs due to the settings of the shutdown facility, each node will become the LEFTCLUSTER state. For details on the settings of the shutdown facility, see "G.2.3.2 Setting Up the Shutdown Facility." For how to recover from the LEFTCLUSTER state, see "6.2 Recovering from LEFTCLUSTER" in "PRIMECLUSTER Cluster Foundation (CF) Configuration and Administration Guide." For the following operations, see "7.2 Operating the PRIMECLUSTER System."
Remove a virtual machine that installs PRIMECLUSTER from targets for cluster functions (such as VMware HA, VMware FT, VMware DRS, and VMware DPM) of VMware.
The following functions are not available in a virtual machine in which PRIMECLUSTER is to be installed.
VMware vMotion
VMware Storage vMotion
Migration with VMware vCenter Converter
Snapshot of VMware
Hot clone
Backup by Data Protection/Data Recovery
Backup by VCB
Set the path policy for the Native Multipathing (NMP) as follows:
When using VMware vSphere 5.1 or earlier
Set to "Most Recently Used".
When using VMware vSphere 5.5 or VMware vSphere 6.0
Set to "Most Recently Used" or "Round Robin".
Settings other than above are not supported.
Following hot swap operations cannot be performed for the virtual machine hardware.
Increasing disk size
Increasing memory
Increasing CPU
Using snapshot
Over committing of memory that causes virtual swap or memory ballooning
Support for multipath software from third parties, contact field engineers.
The following environments and functions are not supported:
iSCSI and FCoE
ESX hosts with different versions
N-Port ID Virtualization (NPIV)
It is necessary to use the hardware version 7 for VMware vSphere 4.x of the virtual machine.
When you cannot refer to a system volume on a disk device in a SAN boot configuration, the PRIMECLUSTER failure detection function cannot be operated depending on the status of the system. In this case, perform a manual switchover by an operator.
Use a shared disk responding to SCSI-3 Persistent Reservation.
When using the file system that is created on the shared disk as Fsystem resources, you need to register all the file systems that are created on the same disk (LUN) or on the same disk class to the same userApplication. Due to the restriction of I/O fencing, you cannot create multiple file systems on one disk (LUN) or on one disk class and register each file system to the different userApplications to monitor and control them.