This document describes how to create the VPP startup configuration
file that should be located located at /etc/vpp/contiv-vswitch.conf
on each node in the cluster with Contiv-VPP.
You need to configure hardware interfaces for use by VPP. First, you
need to find out the PCI address of the host's network interface. On
Debian-based distributions, you can use lshw
:
sudo lshw -class network -businfo
Bus info Device Class Description
========================================================
pci@0000:03:00.0 ens160 network VMXNET3 Ethernet Controller
In our case, it would be the ens3
interface with the PCI address
0000:00:03.0
.
Now, add or modify the VPP startup config file (/etc/vpp/contiv-vswitch.conf
)
to contain the proper PCI address:
unix {
nodaemon
cli-listen /run/vpp/cli.sock
cli-no-pager
coredump-size unlimited
full-coredump
poll-sleep-usec 100
}
nat {
endpoint-dependent
translation hash buckets 1048576
translation hash memory 268435456
user hash buckets 1024
max translations per user 10000
}
acl-plugin {
use tuple merge 0
}
dpdk {
dev 0000:00:03.0
}
api-trace {
on
nitems 5000
}
socksvr {
default
}
statseg {
default
}
Similarly to the single-NIC configuration, use lshw to find out the PCI addresses of all NICs in the system, for example:
$ sudo lshw -class network -businfo
Bus info Device Class Description
====================================================
pci@0000:00:03.0 ens3 network Virtio network device
pci@0000:00:04.0 ens4 network Virtio network device
In the example above, ens3
would be the primary interface and ens4
would
be the interface that would be used by VPP. The PCI address of the ens4
interface would be 0000:00:04.0
.
Now, make sure that the selected interface is shut down, otherwise VPP would not grab it:
sudo ip link set ens4 down
Now, add, or modify the VPP startup config file in /etc/vpp/contiv-vswitch.conf
to contain the proper PCI address:
unix {
nodaemon
cli-listen /run/vpp/cli.sock
cli-no-pager
coredump-size unlimited
full-coredump
poll-sleep-usec 100
}
nat {
endpoint-dependent
translation hash buckets 1048576
translation hash memory 268435456
user hash buckets 1024
max translations per user 10000
}
acl-plugin {
use tuple merge 0
}
dpdk {
dev 0000:00:04.0
}
api-trace {
on
nitems 5000
}
socksvr {
default
}
statseg {
default
}
If assigning multiple NICs to VPP you will need to include each NIC's PCI address
in the dpdk stanza in /etc/vpp/contiv-vswitch.conf
.
Note: On CentOS/RedHat/Fedora distributions, lshw
may not be available
by default, install it by
sudo yum -y install lshw
In regular operation, VPP takes 100% of one CPU core at all times (poll loop).
If high performance and low latency is not required you can "slow-down"
the poll-loop and drastically reduce CPU utilization by adding the following
stanza to the unix
section of the VPP startup config file:
unix {
...
poll-sleep-usec 100
...
}
The power-saving mode is especially useful in VM-based development environments running on laptops or less powerful servers.
To troubleshoot VPP configuration issues in production environments, it is strongly recommended to configure VPP API trace. This is done by adding the following stanza to the VPP startup config file:
api-trace {
on
nitems 500
}
You can set the size of the trace buffer with the attribute.