Network issues - dropped packets

[tjsimmons]

Hi all. I'm very stumped as to what's happening right now. I built a box to run Proxmox on. I have one VM and three lxc containers. Here's the networking configuration.

eth0 - Intel 2-port PCI-e NIC
eth1 - motherboard onboard
eth2 - Intel 2-port PCI-e NIC

vmbr0, 1, 2 are pointed to eth0, 1, 2 respectively.

vmbr0 is my management interface and has a static IP. The other two are assigned via DHCP in the following fashion - vmbr1 is connected to a cable modem, vmbr2 is connected to an 8-port switch.

Untangle is the VM running on this box, doing all of the routing and DHCP serving. It's configuration is eth0 (linked to vmbr1) as a WAN interface, and eth1 (linked to vmbr2) as LAN.

http://imgur.com/a/VcEOs
Linked above are screenshots from a terminal session on my phone (so forgive the keyboard), showing dropped RX packets (740 on both eth0 and eth2, 50 on both vmbr0 and vmbr2) and 152 dropped TX packets on tap100i0.

None of the "physical" devices that the VM and containers think they have show any dropped packets.

I can't tell where the issue is here. The box is at a remote location, and become unresponsive via the Internet for the second time in three days. Only a full reboot (including a reboot of the cable modem) resolved the issue.

Does anyone have any guidance?

 

[manu]

IIRC a small number of dropped network packets as seen on the netstat output is normal (could be ethernet headers checksum mismatch)
Looking at your setup I see that your WAN adresss is obtained via DHCP. Is the cable modem in bridge mode ? ( ie is untangle directly getting its WAN IP adress form your ISP ? )

To debug the issue I would advise you to set a background ping process on untangle to your upstream gateway, to set when the network gets down.

 

[tjsimmons]

IIRC a small number of dropped network packets as seen on the netstat output is normal (could be ethernet headers checksum mismatch)
Looking at your setup I see that your WAN adresss is obtained via DHCP. Is the cable modem in bridge mode ? ( ie is untangle directly getting its WAN IP adress form your ISP ? )

To debug the issue I would advise you to set a background ping process on untangle to your upstream gateway, to set when the network gets down.

Yes - Untangle is getting its WAN address from the ISP.

I've currently got a process running to ping to a service I own to know when the box goes offline. As far as the dropped packets, looking at another nearly identical box I built I have lots of dropped packets across the bridges, so that seems to be normal. What isn't normal are the dropped packets on eth0 and eth2.

The other oddity on this box is three times now, my lxc containers have lost their DHCP addresses.