No Network via bridge

[john.gil]

I've just started using Proxmox and in the KVM Guests such as Windows 2008, Windows 2003, Windows XP, network is not working as long as the ethernet is of bridge type and vmbr0 has been used for it.

I understand that network should be configured inside the Guest like a real server but even after we proceed and set the ip, subnet, gateway and dns entries there is no ping or dns resolution or no website can be surfed by the Guest; once we use NAT things would go fine but through NAT I cannot give the Guest an IP of it's own and use RDP for example.

Please Advise.

 

[tom]

I've just started using Proxmox and in the KVM Guests such as Windows 2008, Windows 2003, Windows XP, network is not working as long as the ethernet is of bridge type and vmbr0 has been used for it.

I understand that network should be configured inside the Guest like a real server but even after we proceed and set the ip, subnet, gateway and dns entries there is no ping or dns resolution or no website can be surfed by the Guest; once we use NAT things would go fine but through NAT I cannot give the Guest an IP of it's own and use RDP for example.

Please Advise.

should work without any issues. pls post your /etc/network/interfaces.

how did you install? do you use the latest version?

 

[john.gil]

Here is the: /etc/network/interfaces

# network interface settings
auto lo
iface lo inet loopback

iface eth0 inet manual

auto vmbr0
iface vmbr0 inet static
    address  64.85.170.88
    netmask  255.255.240.0
    gateway  64.85.160.1
    bridge_ports eth0
    bridge_stp off
    bridge_fd 0

The ip, netmask and the gateway are those provided by the datacenter.

Installation was done through the bare metal installer and we are running Proxmox Virtual Environment 1.3

Thanks

 

[dietmar]

Please use tcpdump to debug that.

 

[john.gil]

Ok, I'll check the tcpdump output, just another question, can a second Guest be configured to use vmbr0 as well? I mean shall one create a seperate bridge for each Guest or many guests can be configured to use vmbr0 as their ethernet device?

 

[tom]

Ok, I'll check the tcpdump output, just another question, can a second Guest be configured to use vmbr0 as well? I mean shall one create a seperate bridge for each Guest or many guests can be configured to use vmbr0 as their ethernet device?

no, you can connect as many as you want.
check this: http://pve.proxmox.com/wiki/Network_Model

 

[john.gil]

Ok, here is the result of tcpdump when using NAT mode actually in Guest I tried to surf the site: http://www.finmyip.com

tcpdump -r tcplog.cap -n | grep -i 72.1.173.139
reading from file tcplog.cap, link-type EN10MB (Ethernet)
07:19:02.943809 IP 64.85.170.xx.59506 > 72.1.173.139.80: S 3508572855:3508572855(0) win 5840 <mss 1460,sackOK,timestamp 18601422 0,nop,wscale 7>
07:19:03.033906 IP 72.1.173.139.80 > 64.85.170.xx.59506: S 1573137973:1573137973(0) ack 3508572856 win 8760 <mss 1460>
07:19:03.033933 IP 64.85.170.xx.59506 > 72.1.173.139.80: . ack 1 win 5840
07:19:03.034449 IP 64.85.170.xx.59506 > 72.1.173.139.80: P 1:372(371) ack 1 win 5840
07:19:03.264257 IP 72.1.173.139.80 > 64.85.170.xx.59506: . ack 372 win 8389
07:19:03.509680 IP 72.1.173.139.80 > 64.85.170.xx.59506: . 1:1461(1460) ack 372 win 8389
07:19:03.509703 IP 64.85.170.xx.59506 > 72.1.173.139.80: . ack 1461 win 8760
07:19:03.509680 IP 72.1.173.139.80 > 64.85.170.xx.59506: P 1461:1892(431) ack 372 win 8389
07:19:03.509714 IP 64.85.170.xx.59506 > 72.1.173.139.80: . ack 1892 win 11680
07:19:03.509682 IP 72.1.173.139.80 > 64.85.170.xx.59506: F 1892:1892(0) ack 372 win 8389
07:19:03.510278 IP 64.85.170.xx.59506 > 72.1.173.139.80: F 372:372(0) ack 1893 win 11680
07:19:03.594549 IP 72.1.173.139.80 > 64.85.170.xx.59506: . ack 373 win 8389
tcpdump: pcap_loop: truncated dump file; tried to read 60 captured bytes, only got 59

72.1.173.139 is the dedicated ip that findmyip.com uses, as you can see both outgoing and reply packets can be seen, as for the second case where vmbr0 is used the output is like this:

tcpdump -r tcplog2.cap -n | grep -i 72.1.173.139
reading from file tcplog2.cap, link-type EN10MB (Ethernet)
08:01:06.212021 IP 64.85.169.aa.1042 > 72.1.173.139.80: S 3856255907:3856255907(0) win 65535 <mss 1460,nop,nop,sackOK>
08:01:09.052726 IP 64.85.169.aa.1042 > 72.1.173.139.80: S 3856255907:3856255907(0) win 65535 <mss 1460,nop,nop,sackOK>
08:01:15.068088 IP 64.85.169.aa.1042 > 72.1.173.139.80: S 3856255907:3856255907(0) win 65535 <mss 1460,nop,nop,sackOK>

I do not see any reply, what would you conclude?

 

[dietmar]

I do not see any reply, what would you conclude?

I would conclude that there is no reply ;-) Or more elaborate, the router does not route packets back to you.

I also do not see any ARP request in the tcp dump, are there any?

- Dietmar

 

[john.gil]

So the next step is to query them why they are blocked at the router level, correct?

There are lots of arps in the form of "who-has IP/Hostname tell Gateway" and icmp packets as well but none that would contain 72.1.173.139, I had to grep the cap file's output to summerize it actually.

let me know if you need to have a look at the complete file.

 

[dietmar]

So the next step is to query them why they are blocked at the router level, correct?

yes

 

[john.gil]

Ok, solved all but something else I see is the node's network is stable but the Guests lose network after a few hours of being up and I often stop them and start again so network comes back, once in a Linux KVM Guest I restarted the machine and network did not come back actually it sould be done via interface to work.

Any ideas?

 

[dietmar]

Do you use virtio? If so, try without.

 

[john.gil]

Not really as on Windows 2003 I could not find drivers for that so we stick to rtl8139

 

[dietmar]

try e1000 with the latest drivers.

 

[SaFuff]

Hello. On the network we have at work, my friend opened my cd drive from his computer. Does anyone know how to do this through cmd.exe?

Thanks

 

[john.gil]

try e1000 with the latest drivers.

and what can be a better option for fully virtualized linux? I have a CentOS that almost loses connection every 24 hours.

 

[john.gil]

I also did a

aptitude update
aptitude safe-upgrade

later I noticed that proxmox advises for the upgrade process only in the form of distro upgrade which should be like

aptitude update
aptitude dist-upgrade

actually I did just update the packages with the first seri of commands, how inconsistent have I made the node by aptitude upgrade? Shall I reload? I remember it also contained a kernel upgrade.