2. Live Migration of VM with SR-IOV VF

2.1. Overview

It is not possible to migrate a Virtual Machine which has an SR-IOV Virtual Function (VF).

To get around this problem the bonding PMD is used.

The following sections show an example of how to do this.

2.2. Test Setup

A bonding device is created in the VM. The virtio and VF PMD’s are added as members to the bonding device. The VF is set as the primary member of the bonding device.

A bridge must be set up on the Host connecting the tap device, which is the backend of the Virtio device and the Physical Function (PF) device.

To test the Live Migration two servers with identical operating systems installed are used. KVM and Qemu 2.3 is also required on the servers.

In this example, the servers have Niantic and or Fortville NIC’s installed. The NIC’s on both servers are connected to a switch which is also connected to the traffic generator.

The switch is configured to broadcast traffic on all the NIC ports. A Sample switch configuration can be found in this section.

The host is running the Kernel PF driver (ixgbe or i40e).

The ip address of host_server_1 is 10.237.212.46

The ip address of host_server_2 is 10.237.212.131

2.3. Live Migration steps

The sample scripts mentioned in the steps below can be found in the Sample host scripts and Sample VM scripts sections.

2.3.1. On host_server_1: Terminal 1

cd /root/dpdk/host_scripts
./setup_vf_on_212_46.sh

For Fortville NIC

./vm_virtio_vf_i40e_212_46.sh

For Niantic NIC

./vm_virtio_vf_one_212_46.sh

2.3.2. On host_server_1: Terminal 2

cd /root/dpdk/host_scripts
./setup_bridge_on_212_46.sh
./connect_to_qemu_mon_on_host.sh
(qemu)

2.3.3. On host_server_1: Terminal 1

In VM on host_server_1:

cd /root/dpdk/vm_scripts
./setup_dpdk_in_vm.sh
./run_testpmd_bonding_in_vm.sh

testpmd> show port info all

The mac_addr command only works with kernel PF for Niantic

testpmd> mac_addr add port 1 vf 0 AA:BB:CC:DD:EE:FF

The syntax of the testpmd command is:

Create bonding device (mode) (socket).

Mode 1 is active backup.

Virtio is port 0 (P0).

VF is port 1 (P1).

Bonding is port 2 (P2).

testpmd> create bonding device 1 0
Created new bonding device net_bond_testpmd_0 on (port 2).
testpmd> add bonding member 0 2
testpmd> add bonding member 1 2
testpmd> show bonding config 2

The syntax of the testpmd command is:

set bonding primary (member id) (port id)

Set primary to P1 before starting bonding port.

testpmd> set bonding primary 1 2
testpmd> show bonding config 2
testpmd> port start 2
Port 2: 02:09:C0:68:99:A5
Checking link statuses...
Port 0 Link Up - speed 10000 Mbps - full-duplex
Port 1 Link Up - speed 10000 Mbps - full-duplex
Port 2 Link Up - speed 10000 Mbps - full-duplex

testpmd> show bonding config 2

Primary is now P1. There are 2 active members.

Use P2 only for forwarding.

testpmd> set portlist 2
testpmd> show config fwd
testpmd> set fwd mac
testpmd> start
testpmd> show bonding config 2

Primary is now P1. There are 2 active members.

testpmd> show port stats all

VF traffic is seen at P1 and P2.

testpmd> clear port stats all
testpmd> set bonding primary 0 2
testpmd> remove bonding member 1 2
testpmd> show bonding config 2

Primary is now P0. There is 1 active member.

testpmd> clear port stats all
testpmd> show port stats all

No VF traffic is seen at P0 and P2, VF MAC address still present.

testpmd> port stop 1
testpmd> port close 1

Port close should remove VF MAC address, it does not remove perm_addr.

The mac_addr command only works with the kernel PF for Niantic.

testpmd> mac_addr remove 1 AA:BB:CC:DD:EE:FF
testpmd> port detach 1
Port '0000:00:04.0' is detached. Now total ports is 2
testpmd> show port stats all

No VF traffic is seen at P0 and P2.

2.3.4. On host_server_1: Terminal 2

(qemu) device_del vf1

2.3.5. On host_server_1: Terminal 1

In VM on host_server_1:

testpmd> show bonding config 2

Primary is now P0. There is 1 active member.

testpmd> show port info all
testpmd> show port stats all

2.3.6. On host_server_2: Terminal 1

cd /root/dpdk/host_scripts
./setup_vf_on_212_131.sh
./vm_virtio_one_migrate.sh

2.3.7. On host_server_2: Terminal 2

./setup_bridge_on_212_131.sh
./connect_to_qemu_mon_on_host.sh
(qemu) info status
VM status: paused (inmigrate)
(qemu)

2.3.8. On host_server_1: Terminal 2

Check that the switch is up before migrating.

(qemu) migrate tcp:10.237.212.131:5555
(qemu) info status
VM status: paused (postmigrate)

For the Niantic NIC.

(qemu) info migrate
capabilities: xbzrle: off rdma-pin-all: off auto-converge: off zero-blocks: off
Migration status: completed
total time: 11834 milliseconds
downtime: 18 milliseconds
setup: 3 milliseconds
transferred ram: 389137 kbytes
throughput: 269.49 mbps
remaining ram: 0 kbytes
total ram: 1590088 kbytes
duplicate: 301620 pages
skipped: 0 pages
normal: 96433 pages
normal bytes: 385732 kbytes
dirty sync count: 2
(qemu) quit

For the Fortville NIC.

(qemu) info migrate
capabilities: xbzrle: off rdma-pin-all: off auto-converge: off zero-blocks: off
Migration status: completed
total time: 11619 milliseconds
downtime: 5 milliseconds
setup: 7 milliseconds
transferred ram: 379699 kbytes
throughput: 267.82 mbps
remaining ram: 0 kbytes
total ram: 1590088 kbytes
duplicate: 303985 pages
skipped: 0 pages
normal: 94073 pages
normal bytes: 376292 kbytes
dirty sync count: 2
(qemu) quit

2.3.9. On host_server_2: Terminal 1

In VM on host_server_2:

Hit Enter key. This brings the user to the testpmd prompt.

testpmd>

2.3.10. On host_server_2: Terminal 2

(qemu) info status
VM status: running

For the Niantic NIC.

(qemu) device_add pci-assign,host=06:10.0,id=vf1

For the Fortville NIC.

(qemu) device_add pci-assign,host=03:02.0,id=vf1

2.3.11. On host_server_2: Terminal 1

In VM on host_server_2:

testpmd> show port info all
testpmd> show port stats all
testpmd> show bonding config 2
testpmd> port attach 0000:00:04.0
Port 1 is attached.
Now total ports is 3
Done

testpmd> port start 1

The mac_addr command only works with the Kernel PF for Niantic.

testpmd> mac_addr add port 1 vf 0 AA:BB:CC:DD:EE:FF
testpmd> show port stats all.
testpmd> show config fwd
testpmd> show bonding config 2
testpmd> add bonding member 1 2
testpmd> set bonding primary 1 2
testpmd> show bonding config 2
testpmd> show port stats all

VF traffic is seen at P1 (VF) and P2 (Bonded device).

testpmd> remove bonding member 0 2
testpmd> show bonding config 2
testpmd> port stop 0
testpmd> port close 0
testpmd> port detach 0
Port '0000:00:03.0' is detached. Now total ports is 2

testpmd> show port info all
testpmd> show config fwd
testpmd> show port stats all

VF traffic is seen at P1 (VF) and P2 (Bonded device).

2.4. Sample host scripts

2.4.1. setup_vf_on_212_46.sh

Set up Virtual Functions on host_server_1

#!/bin/sh
# This script is run on the host 10.237.212.46 to setup the VF

# set up Niantic VF
cat /sys/bus/pci/devices/0000\:09\:00.0/sriov_numvfs
echo 1 > /sys/bus/pci/devices/0000\:09\:00.0/sriov_numvfs
cat /sys/bus/pci/devices/0000\:09\:00.0/sriov_numvfs
rmmod ixgbevf

# set up Fortville VF
cat /sys/bus/pci/devices/0000\:02\:00.0/sriov_numvfs
echo 1 > /sys/bus/pci/devices/0000\:02\:00.0/sriov_numvfs
cat /sys/bus/pci/devices/0000\:02\:00.0/sriov_numvfs
rmmod iavf

2.4.2. vm_virtio_vf_one_212_46.sh

Setup Virtual Machine on host_server_1

#!/bin/sh

# Path to KVM tool
KVM_PATH="/usr/bin/qemu-system-x86_64"

# Guest Disk image
DISK_IMG="/home/username/disk_image/virt1_sml.disk"

# Number of guest cpus
VCPUS_NR="4"

# Memory
MEM=1536

taskset -c 1-5 $KVM_PATH \
 -enable-kvm \
 -m $MEM \
 -smp $VCPUS_NR \
 -cpu host \
 -name VM1 \
 -no-reboot \
 -net none \
 -vnc none -nographic \
 -hda $DISK_IMG \
 -netdev type=tap,id=net1,script=no,downscript=no,ifname=tap1 \
 -device virtio-net-pci,netdev=net1,mac=CC:BB:BB:BB:BB:BB \
 -device pci-assign,host=09:10.0,id=vf1 \
 -monitor telnet::3333,server,nowait

2.4.3. setup_bridge_on_212_46.sh

Setup bridge on host_server_1

#!/bin/sh
# This script is run on the host 10.237.212.46 to setup the bridge
# for the Tap device and the PF device.
# This enables traffic to go from the PF to the Tap to the Virtio PMD in the VM.

# ens3f0 is the Niantic NIC
# ens6f0 is the Fortville NIC

ifconfig ens3f0 down
ifconfig tap1 down
ifconfig ens6f0 down
ifconfig virbr0 down

brctl show virbr0
brctl addif virbr0 ens3f0
brctl addif virbr0 ens6f0
brctl addif virbr0 tap1
brctl show virbr0

ifconfig ens3f0 up
ifconfig tap1 up
ifconfig ens6f0 up
ifconfig virbr0 up

2.4.4. connect_to_qemu_mon_on_host.sh

#!/bin/sh
# This script is run on both hosts when the VM is up,
# to connect to the Qemu Monitor.

telnet 0 3333

2.4.5. setup_vf_on_212_131.sh

Set up Virtual Functions on host_server_2

#!/bin/sh
# This script is run on the host 10.237.212.131 to setup the VF

# set up Niantic VF
cat /sys/bus/pci/devices/0000\:06\:00.0/sriov_numvfs
echo 1 > /sys/bus/pci/devices/0000\:06\:00.0/sriov_numvfs
cat /sys/bus/pci/devices/0000\:06\:00.0/sriov_numvfs
rmmod ixgbevf

# set up Fortville VF
cat /sys/bus/pci/devices/0000\:03\:00.0/sriov_numvfs
echo 1 > /sys/bus/pci/devices/0000\:03\:00.0/sriov_numvfs
cat /sys/bus/pci/devices/0000\:03\:00.0/sriov_numvfs
rmmod iavf

2.4.6. vm_virtio_one_migrate.sh

Setup Virtual Machine on host_server_2

#!/bin/sh
# Start the VM on host_server_2 with the same parameters except without the VF
# parameters, as the VM on host_server_1, in migration-listen mode
# (-incoming tcp:0:5555)

# Path to KVM tool
KVM_PATH="/usr/bin/qemu-system-x86_64"

# Guest Disk image
DISK_IMG="/home/username/disk_image/virt1_sml.disk"

# Number of guest cpus
VCPUS_NR="4"

# Memory
MEM=1536

taskset -c 1-5 $KVM_PATH \
 -enable-kvm \
 -m $MEM \
 -smp $VCPUS_NR \
 -cpu host \
 -name VM1 \
 -no-reboot \
 -net none \
 -vnc none -nographic \
 -hda $DISK_IMG \
 -netdev type=tap,id=net1,script=no,downscript=no,ifname=tap1 \
 -device virtio-net-pci,netdev=net1,mac=CC:BB:BB:BB:BB:BB \
 -incoming tcp:0:5555 \
 -monitor telnet::3333,server,nowait

2.4.7. setup_bridge_on_212_131.sh

Setup bridge on host_server_2

#!/bin/sh
# This script is run on the host to setup the bridge
# for the Tap device and the PF device.
# This enables traffic to go from the PF to the Tap to the Virtio PMD in the VM.

# ens4f0 is the Niantic NIC
# ens5f0 is the Fortville NIC

ifconfig ens4f0 down
ifconfig tap1 down
ifconfig ens5f0 down
ifconfig virbr0 down

brctl show virbr0
brctl addif virbr0 ens4f0
brctl addif virbr0 ens5f0
brctl addif virbr0 tap1
brctl show virbr0

ifconfig ens4f0 up
ifconfig tap1 up
ifconfig ens5f0 up
ifconfig virbr0 up

2.5. Sample VM scripts

2.5.1. setup_dpdk_in_vm.sh

Set up DPDK in the Virtual Machine

#!/bin/sh
# this script matches the vm_virtio_vf_one script
# virtio port is 03
# vf port is 04

/root/dpdk/usertools/dpdk-hugepages.py --show
/root/dpdk/usertools/dpdk-hugepages.py --setup 2G
/root/dpdk/usertools/dpdk-hugepages.py --show

ifconfig -a
/root/dpdk/usertools/dpdk-devbind.py --status

rmmod virtio-pci ixgbevf

modprobe uio
insmod igb_uio.ko

/root/dpdk/usertools/dpdk-devbind.py -b igb_uio 0000:00:03.0
/root/dpdk/usertools/dpdk-devbind.py -b igb_uio 0000:00:04.0

/root/dpdk/usertools/dpdk-devbind.py --status

2.5.2. run_testpmd_bonding_in_vm.sh

Run testpmd in the Virtual Machine.

#!/bin/sh
# Run testpmd in the VM

# The test system has 8 cpus (0-7), use cpus 2-7 for VM
# Use taskset -pc <core number> <thread_id>

# use for bonding of virtio and vf tests in VM

/root/dpdk/<build_dir>/app/dpdk-testpmd \
-l 0-3 -n 4 --socket-mem 350 --  --i --port-topology=chained

2.6. Sample switch configuration

The Intel switch is used to connect the traffic generator to the NIC’s on host_server_1 and host_server_2.

In order to run the switch configuration two console windows are required.

Log in as root in both windows.

TestPointShared, run_switch.sh and load /root/switch_config must be executed in the sequence below.

2.6.1. On Switch: Terminal 1

run TestPointShared

/usr/bin/TestPointShared

2.6.2. On Switch: Terminal 2

execute run_switch.sh

/root/run_switch.sh

2.6.3. On Switch: Terminal 1

load switch configuration

load /root/switch_config

2.6.4. Sample switch configuration script

The /root/switch_config script:

# TestPoint History
show port 1,5,9,13,17,21,25
set port 1,5,9,13,17,21,25 up
show port 1,5,9,13,17,21,25
del acl 1
create acl 1
create acl-port-set
create acl-port-set
add port port-set 1 0
add port port-set 5,9,13,17,21,25 1
create acl-rule 1 1
add acl-rule condition 1 1 port-set 1
add acl-rule action 1 1 redirect 1
apply acl
create vlan 1000
add vlan port 1000 1,5,9,13,17,21,25
set vlan tagging 1000 1,5,9,13,17,21,25 tag
set switch config flood_ucast fwd
show port stats all 1,5,9,13,17,21,25