Remote restart considerations

Edit online

You must understand these considerations before remote restarting a virtual machine.

Storage considerations
  • The storage configuration must match. That is, the number of Virtual I/O Servers and available ports per fabric must be the same on both systems. Additionally, the candidate ports on the target host must be connected to the same set of fabrics that the source virtual machine is connected through.
    For example, if the virtual machine that you want to migrate is connected through a pair of Virtual I/O Servers and each Virtual I/O Server has dual fabric connections, then the destination host must have the following resource configurations:
    • At least one pair of Virtual I/O Servers are storage-ready and are members of the storage connectivity group.
    • Each of these Virtual I/O Servers has a storage-ready Fibre Channel port that is connected to the first fabric and a storage-ready port connected to the second.
    For information about what is required for a Virtual I/O Server to be storage-ready and for a physical Fibre Channel port to be ready, see Storage connectivity groups.
  • When using an imported virtual machine with volumes that use multiple connectivity types (both vSCSI and NPIV), ensure that the right volumes are marked as part of the boot set. To do this, open the virtual machine's details window, click Edit details and select Yes for Boot Set on all appropriate volumes.
  • For NovaLink managed hosts, if the shared storage pool service is running on the host that is down, you must manually switch the service to run on another NovaLink or HMC host. This can be done from the Storage Provider page of the user interface.
  • It is important to prevent two servers from using the same storage at the same time. This is managed in one of these ways:
    • If a storage volume is connected by vSCSI, the volume is removed from the source VIOSs that provide the connection, and the volume is attached to the target host's VIOS or VIOSs.
    • If a storage volume is connected by NPIV, PowerVC checks the fabric to determine whether the WWPNs for the client virtual machine are still active. PowerVC allows a remote restart only if the WWPNs are no longer active.
Management device considerations
  • HMC 850 and later can clean up remote restarted virtual machines from the source host after successful remote restart operations. PowerVC also performs this cleanup, so it is possible that redundant cleanups might happen. In that situation, there might be some harmless error messages in logs and a notification on the PowerVC user interface. You can enable or disable the cleanup capability on the HMC by running this HMC CLI command: rrstartlpar -o set -i "auto_cleanup_enabled=n -r mc", where n=1 to enable cleanup or n=0 to disable cleanup.
  • When NovaLink is the management system, the remote restarted virtual machine is automatically removed from the source host after it comes back up. However, if the virtual machine or the source or destination hosts used in the remote restart operation is removed from PowerVC management before the source host comes up, you will need to manually remove the virtual machine.
Collocation rule considerations
  • Collocation rules are enforced during an untargeted remote restart:
    • If the virtual machine is a member of a collocation rule that specifies affinity, you cannot remote restart it.
    • If the virtual machine is a member of a collocation rule that specifies anti-affinity, you cannot remote restart it to a host that has a virtual machine that is a member of the same collocation rule. Assume the following:
      • VM-1 is on host 1.
      • VM-2 is on host 2.
      • VM-1 and VM-2 are in an anti-affinity collocation rule.
      In the above example, VM-1 cannot be remote restarted to host 2.
  • Collocation rules cannot be enforced during a targeted remote restart. Therefore, virtual machines could be placed on any host and the collocation rule could become violated.
  • Only one migration or remote restart at a time is allowed for virtual machines in the same collocation rule. Therefore, if you try to migrate or remote restart a virtual machine and any other virtual machines in the same collocation rule are being migrated or remote restarted, that request fails.
Image considerations
  • Any virtual optical media on the system is ignored. When the virtual machine is rebuilt, the virtual optical media is not included.
  • If you imported a virtual machine and the virtual machine was deployed by using cloudinit, you must remove - update_hostname from the list of cloud_init_modules of cloud.cfg file at /etc/cloud/cloud.cfg before remote restarting the virtual machine. If you do not remove - update_hostname from the list of modules, cloud-init resets the hostname to the imported virtual machine name value on reboot. See Installing cloud-init on your virtual machine for more details.