I have the following problem.
Two pve's are connected to each other with 10Gbit cards.
MYRICOM Inc. Myri-10G Dual-Protocol NIC (rev 01).
Each card has two ports, these two ports are located in the same vmbr0 bridge on each host.
One of the PVEs has a Debian virtual machine on it, which is acting as a router, vmbr0 is used as the network interface, this Debian also has a 50 VLAN tag bridge on it.
Now the problem itself:
On the second hypervisor we create a virtual machine on Debian, use the same bridge with a 10 gigabit card, specify one of the 50 vlan tags of the debian router as a vlan, set an ip address to this virtual machine and ping the virtual machine from the router - the first 3-6 packets were lost, then 3-9 ping packets successfully returned, then lost again, only now 50+ packets, then again 3-9 packets successfully and so on. But if we ping the Debian router from this virtual machine, the router also pings this machine without problems. If we remove the vlan tag from the vm configuration and leave only the port in the bridge, and in the virtual machine itself create an interface with the previously used tag - no problems.
Also on the Debian router is radvd which annonces certain subnets, vm gets ipv6 address through annoncing and can ping ipv6 internet with no problems.
I decided to investigate tcpdump, thanks to it I found out that debian uses ARP-request of the router with this address, but ARP-response is not received, or rather it receives requests in 40-60 seconds.
To summarize - ipv4 doesn't work for me if I set vlan via pve, but it works if I set it on this virtual machine. At the same time ipv6 works fine either way. Of course, vlan aware is enabled in the bridge configuration. What could be the problem with this? If more information is needed, please ask.
Since my English is not the best, I used a translator, I apologize for logical/punctuation errors.
Two pve's are connected to each other with 10Gbit cards.
MYRICOM Inc. Myri-10G Dual-Protocol NIC (rev 01).
Each card has two ports, these two ports are located in the same vmbr0 bridge on each host.
One of the PVEs has a Debian virtual machine on it, which is acting as a router, vmbr0 is used as the network interface, this Debian also has a 50 VLAN tag bridge on it.
Now the problem itself:
On the second hypervisor we create a virtual machine on Debian, use the same bridge with a 10 gigabit card, specify one of the 50 vlan tags of the debian router as a vlan, set an ip address to this virtual machine and ping the virtual machine from the router - the first 3-6 packets were lost, then 3-9 ping packets successfully returned, then lost again, only now 50+ packets, then again 3-9 packets successfully and so on. But if we ping the Debian router from this virtual machine, the router also pings this machine without problems. If we remove the vlan tag from the vm configuration and leave only the port in the bridge, and in the virtual machine itself create an interface with the previously used tag - no problems.
Also on the Debian router is radvd which annonces certain subnets, vm gets ipv6 address through annoncing and can ping ipv6 internet with no problems.
I decided to investigate tcpdump, thanks to it I found out that debian uses ARP-request of the router with this address, but ARP-response is not received, or rather it receives requests in 40-60 seconds.
To summarize - ipv4 doesn't work for me if I set vlan via pve, but it works if I set it on this virtual machine. At the same time ipv6 works fine either way. Of course, vlan aware is enabled in the bridge configuration. What could be the problem with this? If more information is needed, please ask.
Since my English is not the best, I used a translator, I apologize for logical/punctuation errors.