High CPU usage caused by VVTEP misconfiguration. How do I fix it?
We are quiet new to Arista and currently having this setup and use L3 VXLAN over MLAG:
After moving some SVI from our old network to Arista, we noticed an increase in CPU load to about 80-90%.
MLAG domain 1: LEAF-DC1-01, LEAF-DC1-02
interface Loopback1
ip address 100.64.1.1/32
ip address 100.64.1.100/32 secondary
ip virtual-router mac-address 00:1c:72:00:00:11
interface Vxlan1
vxlan source-interface Loopback1
vxlan virtual-router encapsulation mac-address mlag-system-id
vxlan udp-port 4789
vxlan vlan 1 vni 20001
vxlan vlan 2 vni 20002
vxlan vlan 3 vni 20003
MLAG domain 2: LEAF-DC2-01, LEAF-DC2-02
interface Loopback1
ip address 100.64.2.1/32
ip address 100.64.2.100/32 secondary
ip virtual-router mac-address 00:1c:72:00:00:11
interface Vxlan1
vxlan source-interface Loopback1
vxlan virtual-router encapsulation mac-address mlag-system-id
vxlan udp-port 4789
vxlan vlan 1 vni 20001
vxlan vlan 2 vni 20002
vxlan vlan 3 vni 20003
So the TAC figured out that the high CPU usage was because different VVTEP-IPs on each MLAG which cause loops on ARP sync. Which means we have to change this in MLAG domain 2 so that VVTEP matches the one in MLAG domain 1:
interface Loopback1
no ip address 100.64.2.100/32 secondary
ip address 100.64.1.100/32 secondary
Does anyone know how this will affect traffic?
Also is it neccessary to add
vxlan vtep flood
to interface vxlan 1? How would it look like? Like this?
MLAG domain 1: LEAF-DC1-01, LEAF-DC1-02
interface Vxlan1
vxlan vtep flood 100.64.2.1, 100.64.2.100
MLAG domain 2: LEAF-DC2-01, LEAF-DC2-02
interface Vxlan1
vxlan vtep flood 100.64.1.1, 100.64.2.100