Ask Your Question
0

Fuel: 802.1q fails on non-Admin interface

asked 2015-01-06 19:05:26 -0500

tudor gravatar image

updated 2015-01-07 16:03:23 -0500

I have installed Fuel 6.0 onto a physical server to deploy OpenStack onto a number of physical hosts. In the environment network tab, when I click "Verify networks", Fuel reports that they were not received on any of interfaces associated with Public, Storage, or Management.

My first thought was that it was either a cabling error, or a configuration error. However, I logged into a couple of the hosts and assigned the networks an IP, and was successfully able to ping between these hosts on the correct interfaces.

I ran tcpdump while "Verify networks" was running. I can see what I think are 802.1q packets arrive at the host:

00:51:47.575839 IP 198.18.1.1.31337 > 198.18.1.2.31337: UDP, length 14
    0x0000:  0065 0800 4500 002a 0001 0000 4011 ec9a
    0x0010:  c612 0101 c612 0102 7a69 7a69 0016 0171
    0x0020:  4e61 696c 6775 6e3a 6574 6830 2034
00:51:59.575814 IP 198.18.1.1.31337 > 198.18.1.2.31337: UDP, length 14
    0x0000:  0065 0800 4500 002a 0001 0000 4011 ec9a
    0x0010:  c612 0101 c612 0102 7a69 7a69 0016 0171
    0x0020:  4e61 696c 6775 6e3a 6574 6830 2034

I see also that net-probe left a file at /tmp/net_probe..... that contains:

{"dump_file":"/var/tmp/net-probe-dump","interfaces":{"eth2":"0","eth1":"0","eth0":"101,102"},"ready_address":"127.0.0.1","action":"listen","ready_port":31338}

And /var/tmp/net-probe-dump says:

{
    "eth2": {"0":   {"1": ["eth2"], "3": ["eth2"], "2": ["eth2"], "4": ["eth3"]}}, 
    "eth1": {"0":   {"1": ["eth1"], "3": ["eth1"], "2": ["eth1"]               }}, 
    "eth0": {"101": {"1": ["eth0"], "3": ["eth0"], "2": ["eth0"], "4": ["eth0"]}, 
             "102": {"1": ["eth0"], "3": ["eth0"], "2": ["eth0"], "4": ["eth0"]}}}

This seems to indicate that it's receiving connections from all hosts on all but eth1. What else could be the problem, and what can I do further to debug this?

edit retag flag offensive close merge delete

2 answers

Sort by ยป oldest newest most voted
0

answered 2015-01-14 18:09:00 -0500

tudor gravatar image

It turns out that this was caused either by a faulty network card design or kernel module r8169. The card was a Realtek Semiconductor Co., Ltd. Device [10ec:0123].

It turned out that some packets would occasionally pass, but the majority fail. We noticed that arp packets on startup were consistent, but after that it was pretty random. There are reports that this card works in PCIe 1x slots but not larger ones. We don't have any 1x slots to test with, so we replaced the cards and it worked perfectly.

edit flag offensive delete link more
1

answered 2015-01-13 14:46:48 -0500

dshulyak gravatar image

In your case you should see untagged packets on eth1 and eth2, and 8021q tagged on eth0 (vlan 101 and vlan 102).

Message /var/tmp/net-probe-dump says that current node wasnt able to receive packets from node with id 4. So, please recheck that there is no connection problem on eth1 for 4th node (you can easily find address from output

fuel node

Also it may be possible that there was problems right in time of network verication (i remember couple of such cases), so it wont hurt to restart verification.

If you are sure that network configured correctly, please file a bug with attached diagnostic snapshot.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2015-01-06 19:02:33 -0500

Seen: 251 times

Last updated: Jan 14 '15