Prerequisites for installing a NetScaler VPX instance on Linux-KVM platform

Check the minimum system requirements for a Linux-KVM serves running a NetScaler VPX instance.

CPU requirement:

• 64-bit x86 processors with the hardware virtualization features included in the AMD-V and Intel VT-X processors.

To test whether your CPU supports Linux host, enter the following command at the host Linux shell prompt:

*.egrep'^flags.*(vmx|svm)'/proc/cpuinfo*

If the BIOS settings for the above extension are disabled, you must enable them in BIOS.

• Provide at least 2 CPU cores to Host Linux.

• There is no specific recommendation for processor speed, but higher the speed, the better the performance of the VM application.

Memory (RAM) requirement:

Minimum 4 GB for the host Linux kernel. Add additional memory as required by the VMs.

Hard disk requirement:

Calculate the space for Host Linux kernel and VM requirements. A single NetScaler VPX VM requires 20 GB of disk space.

Software requirements

The Host kernel used must be a 64-bit Linux kernel, release 2.6.20 or later, with all virtualization tools. Citrix recommends newer kernels, such as 3.6.11-4 and later.

Many Linux distributions such as Red Hat, Centos, and Fedora, have tested kernel versions and associated virtualization tools.

Guest VM hardware requirements

NetScaler VPX supports IDE and virtIO hard disk type. The Hard Disk Type has been configured in the XML file, which is a part of the NetScaler package.

Networking requirements

NetScaler VPX supports virtIO para-virtualized, SR-IOV, and PCI Passthrough network interfaces.

For more information about the supported network interfaces, see:

• Provision the NetScaler VPX instance by using the Virtual Machine Manager
• Configure a NetScaler VPX instance to use SR-IOV network interfaces
• Configure a NetScaler VPX instance to use PCI passthrough network interfaces

Source Interface and Modes

The source device type can be either Bridge or MacVTap. In case of MacVTap, four modes are possible - VEPA, Bridge, Private and Pass-through. Check the types of interfaces that you can use and the supported traffic types, as given below.

Bridge:

• Linux Bridge.
• Ebtables and iptables settings on host Linux might filter the traffic on the bridge if you do not choose the correct setting or disable IPtable services.

MacVTap (VEPA mode):

• Better performance than a bridge.
• Interfaces from the same lower device can be shared across the VMs.
• Inter-VM communication using the same
• lower device is possible only if upstream or downstream switch supports VEPA mode.

MacVTap (private mode):

• Better performance than a bridge.
• Interfaces from the same lower device can be shared across the VMs.
• Inter-VM communication using the same lower device is not possible.

MacVTap (bridge mode):

• Better as compared to bridge.
• Interfaces out of same lower device can be shared across the VMs.
• Inter-VM communication using the same lower device is possible, if lower device link is UP.

MacVTap (Pass-through mode):

• Better as compared to bridge.
• Interfaces out of same lower device cannot be shared across the VMs.
• Only one VM can use the lower device.

Note: For best performance by the VPX instance, ensure that the gro and lro capabilities are switched off on the source interfaces.

Properties of source interfaces

Make sure that you switch off the generic-receive-offload (gro) and large-receive-offload (lro) capabilities of the source interfaces. To switch off the gro and lro capabilities, run the following commands at the host Linux shell prompt.

ethtool -K eth6 gro off

ethool -K eth6 lro off

Example:

[root@localhost ~]# ethtool -K eth6

                 Offload parameters for eth6:

                                   rx-checksumming: on

                                   tx-checksumming: on

                 scatter-gather: on

                 tcp-segmentation-offload: on

                 udp-fragmentation-offload: off

                 generic-segmentation-offload: on

                 generic-receive-offload: off

                 large-receive-offload: off

                 rx-vlan-offload: on

                 tx-vlan-offload: on

                 ntuple-filters: off

                 receive-hashing: on

[root@localhost ~]#

Example:

If the host Linux bridge is used as a source device, as in the following example, gro and lro capabilities must be switched off on the vnet interfaces, which are the virtual interfaces connecting the host to the guest VMs.

[root@localhost ~]# brctl show eth6_br

bridge name     bridge  id               STP enabled interfaces

eth6_br         8000.00e0ed1861ae          no         eth6

                                                      vnet0

                                                      vnet2

[root@localhost ~]#

In the above example, the two virtual interfaces are derived from the eth6_br and are represented as vnet0 and vnet2. Run the following commands to switch off gro and lro capabilities on these interfaces.

ethtool -K vnet0 gro off
        ethtool -K vnet2 gro off
        ethtool -K vnet0 lro off
                 ethtool -K vnet2 lro off

Promiscuous mode

The promiscuos mode has to be enabled for the following features to work:

• L2 mode
• Multicast traffic processing
• Broadcast
• IPV6 traffic
• VMAC
• Dynamic routing

Use the following command to enable the promicuous mode.

[root@localhost ~]# ifconfig eth6 promisc
[root@localhost ~]# ifconfig eth6
eth6       Link encap:Ethernet  HWaddr 78:2b:cb:51:54:a3
          inet6 addr: fe80::7a2b:cbff:fe51:54a3/64 Scope:Link
          UP BROADCAST RUNNING PROMISC MULTICAST  MTU:9000  Metric:1
          RX packets:142961 errors:0 dropped:0 overruns:0 frame:0
          TX packets:2895843 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:14330008 (14.3 MB)  TX bytes:1019416071 (1.0 GB)

[root@localhost ~]#

Module required

For better network performance, make sure the vhost_net module is present in the Linux host. To check the existence of vhost_net module, run the following command on the Linux host :

lsmod | grep "vhost\_net"

If vhost_net is not yet running, enter the following command to run it:

modprobe vhost\_net