Abstract
| Goal | Separate network traffic into multiple networks on one or more interfaces to improve the performance and security of Red Hat Virtualization. |
| Objectives |
|
| Sections |
|
| Lab |
Managing RHV Networks |
After completing this section, you should be able to create logical networks to segregate traffic in a data center.
Network configuration is one of the most important factors influencing the performance of your virtualization environment. Networking in Red Hat Virtualization is defined in several layers. The underlying physical networking infrastructure must exist and be configured for connectivity between hardware and the logical components of the RHV environment.
You create logical networks to segregate different types of network traffic onto separate VLANs or physical networks for improved security and performance. For example, separate VLANs can be used for management traffic, storage traffic, or to isolate the traffic of a set of virtual machines.
Logical Networks
Logical networks are defined in a data center and assigned to one or more clusters. A logical network can be assigned to multiple clusters in a data center, to provide communication between VMs in different clusters.
New logical networks are defined with a unique name, the data center in which it resides, and whether it will be used for workload (VM network) traffic. Logical networks require setting a unique VLAN tag (VLAN ID) if this virtual network will share access with any other virtual network on a RHV host physical NIC. If needed, you can also configure Quality of Service (QoS) and bandwidth limiting settings for the logical network.
Designating VM Networks
Logical networks are either designated as VM network or else they are regarded as infrastructure networks. A VM network is a logical network that will connect to virtual network interface cards (vNICs) to carry virtual machine application traffic. A software-defined Linux bridge is created, per logical network, on the RHV host to provide connectivity between the host's physical NIC and the virtual machine vNICs configured to use that logical network. Linux bridge performance is proportional to that of the host's physical NIC, and limited by any QoS settings.
A data center and its clusters can have multiple VM networks, for traffic management and security reasons. When virtual machines are created, each vNIC is assigned to a VM network.
All other logical networks are regarded as infrastructure networks for communication between RHV-M and RHV host only, and are not connected to virtual machines. Because infrastructure networks do not connect to VMs they do not need an associated Linux bridge on RHV hosts.
Infrastructure networks must be configured at the cluster level to indicate what type of traffic it carries. Each host in the cluster also must have the host's correct physical network interface configured for that network. The logical network types are discussed next.
Types of Logical Networks
Logical networks can be configured to segregate different traffic types on different logical networks.
The initial, single default logical network, called ovirtmgmt, is configured as both a VM network and also to handle all infrastructure traffic.
By default, ovirtmgmt is used for management, display and migration network traffic, in addition to VM traffic.
Although this a functional configuration, it provides no boundaries between different network traffic types.
Red Hat recommends that you plan and create additional logical networks to segregate traffic. The following lists provides examples of practical traffic segregation.
Segregating Network Traffic by Types
This network role facilitates VDSM communication between the RHV-M and the RHV hosts.
By default, it is created during the RHV-M engine deployment and named ovirtmgmt.
It is the only logical network created automatically; all others are created according to environment requirements.
This network role is assigned to a network to carry the virtual machine display (SPICE or VNC) traffic from the Administration or VM Portal to the host running the VM. The RHV host then accesses the VM console using internal services. Display networks are not connected to virtual machine vNICs.
Any logical network designated as a VM network carries network traffic relevant to the virtual machine network. This network is used for traffic created by VM applications and connects to VM vNICs. If applications require public access, this network must be configured to access appropriate routing and the public gateway.
A storage network provides private access for storage traffic from RHV hosts to storage servers. Multiple storage networks can be created to further segregate file system based (NFS or POSIX) from block based (iSCSI or FCoE) traffic, to allow different performance tuning for each type. Jumbo Frames are commonly configured on storage networks. Storage networks are not a network role, but are configured to isolated storage traffic to separate VLANs or physical NICs for performance tuning and QoS. Storage networks are not connected to virtual machine vNICs.
This network role is assigned to handle virtual machines migration traffic between RHV hosts. Assigning a dedicated non-routed migration network ensures that the management network does not lose connection to hypervisors during network-saturating VM migrations.
This network role is assigned to provide traffic from Red Hat Gluster Servers to GlusterFS storage clusters.
Although not a network role, creating a network for isolating fencing requests ensure that this critical requests are not missed. RHV-M does not perform host fencing itself but sends fence requests to the appropriate host to execute the fencing command.
Required and Optional Networks
When created, logical networks may be designated as Required at the cluster level. By default, new logical networks are added to clusters as required networks. Required networks must be connected to every host in the cluster, and are expected to always be operational.
When a required network becomes nonoperational for a host, that host's virtual machines are migrated to another cluster host, as specified by the current cluster migration policy. Mission-critical workloads should be configured to use required networks.
Logical networks that are not designated as required are regarded as optional. Optional networks may be implemented only on the hosts that will use them. The presence or absence of optional networks does not affect the host's operational status.
When an optional network becomes nonoperational for a host, that host's virtual machines that were using that network are not migrated to another host. This prevents unnecessary overhead caused by multiple, simultaneous migrations for noncritical network outages. However, a virtual machine with a vNIC configured for an optional VM network will not start on a host that does not have that network available.
Logical network configuration occurs at each layer of the RHV environment.
Logical networks are defined at the data center level. Each data center has the ovirtmgmt management network by default. Additional logical networks are optional but recommended. VM network designation and a custom MTU are set at the data center level. A logical network defined for a data center must be added to the clusters that use the logical network.
Logical networks are available from the data center, and added to clusters that will use them. Each cluster is connected to the management network by default. You can add any logical networks to a cluster if they are defined for the parent data center. When a required logical network is added to a cluster, it must be implemented on each cluster host. Optional logical networks can be added to hosts as needed.
Virtual machine logical networks are connected to each host in a cluster and implemented as a Linux bridge device associated with a physical network interface. Infrastructure networks do not implement Linux bridges but are directly associated with host physical NICs. When first added to a cluster, each host has a management network automatically implemented as a bridge on one of its NICs. All required networks in a cluster must be associated with a NIC on each cluster host to become operational for the cluster.
Logical networks that are available for a host are available to attach to a virtual machine NIC on that host. The virtual machine then gains access to other systems and destinations available on the logical network through the connected vNIC.
Performance Considerations
Gigabit Ethernet is sufficient for the management network, and is typically sufficient for the display network. Any migration and storage networks you add will perform better as dedicated high-bandwidth networks or VLANs. Use 10 GbE or 40 GbE infrastructure when available. Smaller networks can be aggregated larger throughput by using network bonding or teaming. Bandwidth requirements for VM Networks must be calculated from your application requirements.
Red Hat recommends using larger networks integrated with VLANs and advanced QoS features to more easily manage host infrastructure and networking performance of the RHV environment.
Create logical networks in the page under the menu. Click to open a New Logical Network window. Select the data data center for the network and assign a unique name. You need to know if the network will using a VLAN tag, which should already have been planned across the RHV environment. Click Enable VLAN tagging and enter the VLAN ID number assigned. If this network is to be use for virtual machine traffic, select VM network. Click on , to attach the new network to all data center clusters as a required network.
To select a cluster, click on then . Highlight a cluster in the list and click on it. In the upper pane that displays, select the Logical Networks tab.
Using the button, access the Manage Networks window. Using the available check boxes, specify which network will carry each type of infrastructure traffic. You can also assign or unassign networks to the cluster, and designate whether each is required or optional.
Finally, the logical network must be attached to a physical network interface on one or more hosts in the cluster. This topic is covered in the next section of this chapter.
Further information is available in the Logical Networks chapter of the Administration Guide for Red Hat Virtualization; at https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.3/html-single/administration_guide/index#chap-Logical_Networks
Further information is available in the Network chapter of the Technical Reference for Red Hat Virtualization; at https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.3/html-single/technical_reference/index#chap-Network