Skip to content

Latest commit

 

History

History
166 lines (83 loc) · 5.46 KB

Cluster.rst

File metadata and controls

166 lines (83 loc) · 5.46 KB

Clustering

Nodes running MCVirt can be joined together in a cluster - this allows the synchronization of VM/global configurations.

Viewing the status of a cluster

To view the status of the cluster, run the following on an MCVirt node:

mcvirt info

This will show the cluster nodes, IP addresses, and status.

Adding a new node

It is best to join a blank node (containing a default configuration without any VMs) to a cluster.

When a new node is connected to a cluster, the configuration from the present nodes in the cluster (e.g. users, permissions, networks etc.) are pushed to the new node and any existing configuration is replaced.

Note: Always run the mcvirt cluster add command from the source machine, containing VMs, connecting to a remote node that is blank.

The new node must be configured on separate network/VLAN for MCVirt cluster communication.

The IP address that MCVirt clustering/DRBD communications will be performed over must be configured by performing the following on both nodes:

mcvirt node --set-ip-address <Node cluster IP address>

This configuration can be retrieved by running mcvirt info.

Joining the node to the cluster

Note: The following can only be performed by a superuser.

  1. From the remote node, run:
mcvirt cluster get-connect-string

The connect string will be displayed

  1. From the source node, run:
mcvirt cluster add-node --connect-string <connect string>

where <connect string> is the string printed out in step 1.

  1. The local node will connect to the remote node, ensure it is suitable as a remote node, setup authentication between the nodes and copy the local permissions/network/virtual machine configurations to the remote node. Note: All existing data on the remote node will be removed.

Removing a node from the cluster

Note: The following can only be performed by a superuser.

To the remove a node from the cluster, run:

mcvirt cluster remove-node --node <Remote Node Name>

Get Cluster information

  • In order to view status information about the cluster, use the 'info' parameter for MCVirt, without specifying a VM name:

    mcvirt info

Virtual machine migration

  • VMs that use DRBD-based storage can be migrated to the other node in the cluster, whilst the VM is powered off, using:

    mcvirt migrate --node <Destination node> <VM Name>

  • Additional parameters are available to aid the migration and minimise downtime:

    • --wait-for-shutdown, which will cause the migration command to poll the running state of the VM and migrate once the VM is in a powered off state, allowing the user to shutdown the VM from within the guest operating system.
    • --start-after-migration, which starts the VM immediately after the migration has finished
    • --online, which will perform online migration. Note: these cannot be used with either of the previous arguments.

DRBD

DRBD is used by MCVirt to use replicate storage across a 2-node cluster.

Once DRBD is configured and the node is in a cluster, 'DRBD' can be specified as the storage type when creating a VM, which allows the VM to be migrated between nodes.

Configuring DRBD

  1. Ensure the package drbd8-utils is installed on both of the nodes in the cluster

  2. DRBD data will be transmitted over the 'cluster' address. Ensure that this has been set and that the network is segemneted from other network traffic (e.g. by using VLANs).

  3. Perform the following MCVirt command to configure DRBD:

    mcvirt drbd enable

DRBD verification

MCVirt has the ability to start/monitor DRBD verifications (See the DRBD documentation).

The verification can be performed by using:

mcvirt verify <--all>|<VM Name>

This will perform a verification of the specified VM (or all of the DRBD-backed VMs, if '--all' is specified). Once the verification is complete, an exception is thrown if any of the verifications fail.

The status of the latest verification is captured and will stop users from starting/migrating the VM.

If the verification fails:

  • The DRBD volume can be resynced using resync:

    mcvirt resync --source-node=<Node>|--auto-determine <VM Name>

Troubleshooting

Failures during VM creation/deletion

When a VM is created, the following order is performed:

  1. The VM is created, configured with the name, memory allocation and number of CPU cores
  2. The VM is then created on the remote node
  3. The VM is then registered with LibVirt on the local node
  4. The hard drive for the VM is created. (For DRBD-backed storage, the storage is created on both nodes and synced)
  5. Any network adapters are added to the VM

If a failure of occurs during steps 4/5, the VM will still exist after the failure. The user should be able to see the VM, using mcvirt list.

The user can re-create the disks/network adapters as necessary, using the mcvirt update command, using mcvirt info <VM Name> to monitor the virtual hardware that is attached to the VM.