upgrade from Ubuntu Server 8.04 to 10.04 breaks network

[NewtownGuy]

I upgraded two Intel D945GCLF2 machines from Ubuntu Server 8.04 LTS to 10.04 LTS. I upgraded BIOS to the latest version (April 2010), set the size of the Internal Graphics Display (IGD) memory from 128 MB to 256 MB, and installed all the latest updates to 8.04, as recommended in several forums. Unfortunately, neither machine has been able to connect to the LAN since the OS upgrade.

I copied all /var/log/dist-upgrade/ files to another machine before rebooting the upgraded machine, so I can provide those files if there are questions about the procedure.

Can someone help me fix these machines ?

Thank you in advance.

-- NewtownGuy

 

[Noway2]

The first thing to look at would be the output of ifconfig. This will show if the interfaces are "up" and whether or not they have been configured. If not, look in /var/log/syslog and and or dmesg and see if you have any error messages. Then try executing ifup <interface> (such as ifup eth0 and or /etc/init.d/networking restart and look for error messages. You will need to run these with sudo.

The second place to look would be in the /etc/network/interfaces. Are these machines statically configured or dynamic?

When you say "connect to the lan" what symptoms are you experiencing. The description says unable to get a DHCP address to me and I would like to clarify if you mean something different.

If you machines get or are assigned a valid IP address, can you ping another host on the network? If not, what error message do you receive?

Also, look at the output of your routing table (assuming you get an IP? (enter command route) One key point is that you can't have multiple default hosts, which can be a problem if you have dual NIC cards. The routing table will show you this.

 

[NewtownGuy]

to: Noway2

Thank you for your reply.

The machines do not get an IP address. ifconfig does not show any network connection except for loopback. It is as if the NIC (there is only one) does not exist. The machine, prior to being upgraded, had a simple /etc/network/interfaces with merely the loopback and one DHCP reference.

The NIC on these particular mainboards was a problem in Ubuntu Server 8.04 because it was not recognized correctly by the installer. That was fixed by a script and worked fine. I did a fresh install on a third, identical machine with Ubuntu Server 10.04, and it is OK. So something in the upgrade from 8.04 to 10.04 must be breaking the NIC.

Is there any place else to look ? These servers were upgraded via the network and are hard to touch, especially now that they can't use the network.

-- NewtownGuy

 

[Noway2]

Sorry about making a pun, but I would suggest looking in the upgrade and installation issues section of Ubuntuforums.org. It is possible that someone has faced this or a similar issue. I suggest 'searching' for your make/model of ethernet card.

If a fresh install worked fine, my hunch is that something went wrong with a merge or over-write of one of the configuration files. Every time I have performed a distrobution upgrade, I have had problems with changes in the configuration files.

You said that you ran a script initially to get it working. What was the script and have you tried running it again?

One thing to look into is that the default installation may have added some proprietary binary only drivers and you may either be able to make use of them or be having a problem with them.

 

[NewtownGuy]

To: Noway2

Thank you for your input.

The D945GCLF2 has a built-in NIC that was not recognized correctly by Ubuntu Server 8.04. This was fixed by the switchmods script that handles the mixup between the R8168 and R8169 NIC's. This script, which includes a compilation step, will not run on Ubuntu Server 10.04, probably because a required library is no longer present.

Prior to doing the upgrade, I found a forum that described the use of this mainboard, but it did not mention the NIC problem. The forum was a development forum, and something in the Ubuntu 10.04 installer must have changed since then.

I am not able to check /etc/network/interfaces at the moment, but it is the first place I will look once I get my hands on the now, out-of-service machines. That forum clued me into the problem with the memory size for the graphics display, and the need to update BIOS. However, based on my experience with two machines, updating the BIOS had no effect, but increasing the memory size did.

What other files should I check to see if the upgrade damaged something ?

--

 

[BadBigBen]

Question:

have you ever attempted to install the Linux Driver that is offered on the RealTek website?

Ben
"If it works don't fix it! If it doesn't use a sledgehammer..."
How to ask a question, when posting them to a professional forum.
Only ask questions with yes/no answers if you want "yes" or "no"

 

[NewtownGuy]

To: BigBadBen

Thank you for your suggestion.

I am using a Realtek r8168 driver for the 2.6 kernal, but not the latest one.

Does a fresh 10.04 install use the latest driver, whereas the older driver is not compatible with an upgrade from 8.04 to 10.04 ?

--

 

[Noway2]

That is a good question. I would think that an upgrade would also upgrade the drivers. Also, one of the key differences in Linux is that the drivers are loaded at boot up, not during the installation process.

I was answering a similar condition in another forum, where someone was having problems with Slackware not recognizing their drivers. I came across an Ubuntu forum post that suggested a couple of diagnostic commands. I have copied the relevant information from that post here and maybe it will give you some more insight.

link to other thread

Have a look at this thread. It suggests a couple of network diagnostic tools that I haven't used (or seen before). Specifically it recommends running lshw -C network, mii-tool -v or ethtool as these will tell you what the OS 'sees' as far as the adapter. There is also a link to drivers for the card. These might give you some more diagnostic information. One thing I once ran into on Slackware, with Zenwalk being in many ways similar, is that the generic kernel did not automatically load all of the modules necessary for the hardware. It was necessary to run a script that created an ram image that was loaded at boot with these modules.

 

[Noway2]

I just had an idea for you. If a clean installed worked on the other machine and you think it might work on this one, you could backup your /home directories, make a list of all the packages that you currently have installed, do an install of the new version, re-install all your packages, then restore your home directory. This should get you 'moved' over to the new system. I used this technique to migrate a system once.

Here are the commands to backup and restore the package list, courtesy of John Broome of TriLUG.org. (type the stuff between the << and >>)

To replicate your packages selection on another machine (or restore it if re-installing), you can type << aptitude --display-format '%p' search '?installed!?automatic'
> ~/my-packages >>, move the file "my-packages" to the other machine, and there type << sudo xargs aptitude --schedule-only install < my-packages ; sudo aptitude install >>

 

[NewtownGuy]

To: Noway2

Thank you for your many suggestions.

Let me tell you what the REAL problem is...

The machine that is not working is 3,000 miles away and at the customer's location. At this point, he will probably have to return the machine to me to fix it. Before updating that machine via the Internet, I updated two other machines successfully via the network. However, I did not realize that the first two machines used a different mainboard than the third one until there was an upgrade problem.

I've been hoping there is a simple set of instructions that I can give the customer to fix the machine so he does not have to send it back to me. I have not asked him to execute the test commands you gave on Jan 24 because I want to explore all options before asking him to try again.

By the way, ethtool needs an option, I'm not sure which one.

--

 

[BadBigBen]

ethtool -i|--driver eth0


see:

http://linux.die.net/man/8/ethtool

Ben
"If it works don't fix it! If it doesn't use a sledgehammer..."
How to ask a question, when posting them to a professional forum.
Only ask questions with yes/no answers if you want "yes" or "no"