My question revolves around traffic flow in Proxmox. I am trying to understand if it internally "routes" traffic between subnets/vlans internally that it "knows" about.
Take the simple case (ignoring the image).
If I have VLAN100/10.0.100.0/24 and VLAN200/10.0.200.0/24 on my router and switches. I trunk VLAN100/VLAN200 down into Proxmox and put them on a MyVM1 (VLAN100/10.0.100.10) and MyVM2 (VLAN200/10.0.200.20). MyVM1 tries to ping MyVM2. Since Proxmox is "aware" of VLAN100/VLAN200 and the 10.0.100.0/24 and 10.0.200.20/24 subnets will it self "route" the traffic or will that traffic go up to the router and back down to the other VM?
For a more specific use case.
I am dealing with an external Ceph Cluster that isn't internal to proxmox. Ceph Public/Frontend traffic sits on VLAN20. If I put "storage" traffic on VLAN20 and VM Client traffic on VLAN30, and the VM needs to access some other resources through Ceph Frontend (S3 for example), since VLAN20 and VLAN30 cannot communicate with each other directly the VM client traffic would be limited to 1G, since that traffic would have to go "up" to the the router and back down to the Ceph public network.
Basically the two options if VM (VLAN30) tries to communicate with (VLAN20)
1) Proxmox "internally" routes VM traffic from VLAN30 to VLAN20
2) Proxmox sends VM traffic up to VLAN30 up to router, router routes and sends traffic down to VLAN20
I would be confident that it would be #2 if it was just two networked devices, but I'm unsure on if Proxmox is doing some type of magic.
This question is stemming from wanting to separate VM client traffic, storage traffic, migration/replication traffic. The conclusion I am coming to is I can't because everything would need to communicate with the Ceph front end network and it would have to be routed to cross VLANs/Subnets which would reduce bandwidth down to 1G.
Take the simple case (ignoring the image).
If I have VLAN100/10.0.100.0/24 and VLAN200/10.0.200.0/24 on my router and switches. I trunk VLAN100/VLAN200 down into Proxmox and put them on a MyVM1 (VLAN100/10.0.100.10) and MyVM2 (VLAN200/10.0.200.20). MyVM1 tries to ping MyVM2. Since Proxmox is "aware" of VLAN100/VLAN200 and the 10.0.100.0/24 and 10.0.200.20/24 subnets will it self "route" the traffic or will that traffic go up to the router and back down to the other VM?
For a more specific use case.
I am dealing with an external Ceph Cluster that isn't internal to proxmox. Ceph Public/Frontend traffic sits on VLAN20. If I put "storage" traffic on VLAN20 and VM Client traffic on VLAN30, and the VM needs to access some other resources through Ceph Frontend (S3 for example), since VLAN20 and VLAN30 cannot communicate with each other directly the VM client traffic would be limited to 1G, since that traffic would have to go "up" to the the router and back down to the Ceph public network.
Basically the two options if VM (VLAN30) tries to communicate with (VLAN20)
1) Proxmox "internally" routes VM traffic from VLAN30 to VLAN20
2) Proxmox sends VM traffic up to VLAN30 up to router, router routes and sends traffic down to VLAN20
I would be confident that it would be #2 if it was just two networked devices, but I'm unsure on if Proxmox is doing some type of magic.
This question is stemming from wanting to separate VM client traffic, storage traffic, migration/replication traffic. The conclusion I am coming to is I can't because everything would need to communicate with the Ceph front end network and it would have to be routed to cross VLANs/Subnets which would reduce bandwidth down to 1G.
Attachments
Last edited:
