Fedora 9 makes confusion with 2 NIC

[pbonavoglia]

I have installed Fedora 9 on a server HP Proliant ML 350 G5.
The server has 2 NIC as reported here:

# lspci | grep Ethernet
03:00.0 Ethernet controller: Broadcom Corporation NetXtreme II BCM5708 Gigabit Ethernet (rev 12)
14:01.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8169 Gigabit Ethernet (rev 10)

One is linked to an ADSL router connected to internet, the other to the LAN of my school.

During installation the two NICs were recognized normally, but on first run one was lost and I had to install it manually. After that everything worked fine.

But when I reboot the server crazy things happen: sometimes the server doesn't see the 2 NICs and ifconfig reports only the loopback IF. Next boot and the server sees only one NIC (no clear rule, sometimes the Broadcom, sometimes the Realtek) and the ifconfig reports one only; the other (sometimes eth0, sometimes eth1 has disappeared); lspci reports both as seen above.

I tried to create a modprobe.conf file with 2 alias
alias eth0 r8169
alias eth1 bnx2

and restarted the network service; it worked! The 2 NICs were again working ... but at next reboot things went even worse: the boot hanged on eth0 for many minutes and I had to turn off the server and to use a LIVE distribution to empty again the modprobe.conf file. Having done so the server restarted but again seeing only one NIC (or none). There is no way to see both NICs.

Ok it's becoming a nightmare!

Can someone help me?

PS Another "minor" problem: the server has no floppy disk and yet Fedora at every boot loses 3 minutes looking for a floppy controller. Is there a way to tell Fedora there is no floppy here?

Thanks to anyone who can help.

Paolo Bonavoglia

[pbonavoglia]

I had no reply to my message and the nightmare goes on!

I give some more detail abot, hoping to get some help.

Fedora gives this message during boot

udevd-event[823]: run_program: '/sbin/modprobe' abnormal exit

then behaviour seems to become completely unpredictable!

Some times Fedora 9 swaps eth0 with eth1

then

Some times Fedora 9 doesn't see any network card.
Some times it sees one network card, but not the other.

Sometimes it sees both cards, but this success is by no way persistent; next time You get one of the preceding!



Paolo Bonavoglia

[marko]

Run system-config-network, then use the Edit button for both interfaces to get to the Hardware Device tab and in there use the checkbox --> "Bind to MAC address". This way eth0 and eth1 will be bonded to their proper MAC addresses so Fedora won't mix them up.

When this is correct, the ifcfg-eth0 and ifcfg-eth1 files in /etc/sysconfig/network-scripts should have HWADDR values set
to those MACs

[pbonavoglia]

Originally Posted by marko

Run system-config-network, then use the Edit button for both interfaces to get to the Hardware Device tab and in there use the checkbox --> "Bind to MAC address". This way eth0 and eth1 will be bonded to their proper MAC addresses so Fedora won't mix them up.

When this is correct, the ifcfg-eth0 and ifcfg-eth1 files in /etc/sysconfig/network-scripts should have HWADDR values set
to those MACs

I have tried that many times, I tried this morning again.

Nothing to do.

When I reboot it's the same mess. Now I see the broadcom card but not the realtek (PCI). Tried to bind the card to the MAcAddr, checked the 2 files ifcfg, restart the network, but ifconfig reports only one card.

Paolo Bonavoglia

[stevea]

WTF do you mean when you claim a NIC
"one was lost and I had to install it manually"
"Now I see the broadcom card but not the realtek"
"sometimes eth0, sometimes eth1 has disappeared"
????

We have NO IDEA what you think "lost" and "disappear" and "install" means.

lspci examines the PCI bus and recognizes the vendor and product IDs. If lspci "finds" the *INTERFACE* this just means the hardware is present and responding to basic accesses.

"lspci -k" also reports the software driver installed for the device.


"ifconfig" reports enet INTERFACES that have drivers correctly and are configured into an UP state.
"ifconfig -a" reports enet INTERFACES that have drivers correctly installed (UP or DOWN state)

PLEASE report exaclty what you are using as a test. When you say "disappear" it could mean
1/ thehardware is flakey,
2/ the driver is not installed,
3/ the wrong driver is installed, or
4/ the correct driver is installed, but the interface is DOWN.

Which is it ?


What do you mean "install it manually"? Did you reseat the card ? Did you install a driver ? Did you config the interface to an UP state ?

You have supplied little solid information - so no answer is possible.

Gobbledegook!

[pbonavoglia]

Originally Posted by stevea

PLEASE report exaclty what you are using as a test. When you say "disappear" it could mean
1/ thehardware is flakey,
2/ the driver is not installed,
3/ the wrong driver is installed, or
4/ the correct driver is installed, but the interface is DOWN.

Which is it ?

What do you mean "install it manually"? Did you reseat the card ? Did you install a driver ? Did you config the interface to an UP state ?

OK I understand I didn't give enough information but there would be such a lot of details, my mail would have been awfully long.

In fact at every reboot the NIC "disappeared" (one or both) in many different ways:

d1. Disappear from ifconfig.
d2. Disappear from the NIC list in the system-config-network or be present in the list but refuse to be enabled (UP state) with messages like: "There is no eth0 (or eth1) present"
d3. Disappear from the hardware list in the system-config-network (worst of all; I couldn't do anything but reboot once again)
d4. Disappear from ifconfig and the rest as eth1, while reappearing with other names, last time dev3167, even in the ifcfg-eth1 file (DEVICE=dev3167)
d5. While booting the system hanged while triyng to install eth0 or eth1, and You have to turn off the server

By lspci the 2 NIC are always present.
While booting You see several error messages, in particular (always) the "/sbin/modprobe abnormal exit"
Worst of all: when You have configured the system working good, next reboot will again produce one of the above "disappearing" and You have to condifure all over again (when possible).


When I say "install manually" I intend different ways too:

a1. (for d1 and d2) With the system.config-network tool delete, if present, the NIC from the list and then create new binding to the right hardare and then assigning the right IP.
a2. (for d4) assigning IP to the new name (e.g. dev3167) with ifconifg or network config tool.
a3 write a new modprobe.conf (alias eth0 bnx2 alias eth1 r8169) and then restarting the network service (not always working, seems better to leave modprobe.conf empty).

Last thing: the HP Proliant has a third NIC reserved for ILO remote control; apparently nothing to do with this problem, but seems that if You disable the NIC on the ILO config tool things go worse (usually both NIC disappear from ifconfig and system-config-network)

Hope all these details will help.

I installed Fedora 1, 3, 8 on other PC and servers and had absolutely no problem. I wonder if it may be the case to try another O.S. e.g CentOS?

Paolo Bonavoglia

[stevea]

Originally Posted by pbonavoglia

OK I understand I didn't give enough information but there would be such a lot of details, my mail would have been awfully long.

Long isn't a problem. Lack of detail means we all waste time.

In fact at every reboot the NIC "disappeared" (one or both) in many different ways:

d1. Disappear from ifconfig.
d2. Disappear from the NIC list in the system-config-network or be present in the list but refuse to be enabled (UP state) with messages like: "There is no eth0 (or eth1) present"
d3. Disappear from the hardware list in the system-config-network (worst of all; I couldn't do anything but reboot once again)
d4. Disappear from ifconfig and the rest as eth1, while reappearing with other names, last time dev3167, even in the ifcfg-eth1 file (DEVICE=dev3167)
d5. While booting the system hanged while triyng to install eth0 or eth1, and You have to turn off the server

d1 - do you mean "ifconfig" or that AND "ifconfig -a" ?
d2 - meaningless unless you give the output from "ifconfig -a" or "ip addr".
d3 - *sounds* like the driver isn't loaded, which is why you must look through "dmesg" output when this happens.
d4 - refer to d1.
d5 - can't debug if it hangs.

So far you are batting ZERO for providing the information I expected.

When I say "install manually" I intend different ways too:

Like the red queen - words mean whatever you want - eh ? BTW those are INTERFACES not NICs, NICs are hardware.

I suggest you get advice from a Linux guru with ESP. No one can debug from this confused incomplete description.

[pbonavoglia]

Well now the problem is over.

To make it a long story, and be patient with my English, it's not my mother language, I will remember a problem I had s few weeks ago; Telecom Italia had installed a new ADSL line and a new router in our new lab and the line didn't work or worked in an erratic way. I called Telecom support and I talked to a person, let's call A.
I explained the problem and A said: 'You must be clearer, what do You mean with this, what do You mean with that?".
I gave more details and A sai: "But You are still not very clear, what do You mean with this, what do You mean with that?".
I gave even more details and A sai: "But You are still not very clear, what do You mean with this, what do You mean with that?"
The game could go on forever. Luckily I finally managed to talk to another Teleoom person, let's call B.
I restated our problem in very few words and in about three seconds B understood that the router had been misconfigured by Telecom (something I was triyng to explain to A), Telecom reconfigured the router and the problem was over.
What to say about A's behaviour?

Oooh I guess he was appliying the good old rule: "If You don't understand the problem of the client ... blame the client!"

Of course nothing to do with this problem of 2 NIC's.

Anyway coming to my problem with Fedora 9 that made a mess with eth0 and eth1 at every reboot.
Yestersay I disintalled Fedora and installed CentOS: things went quite better, no more chaos of error messages at reboot, but still the eth1 card worked erratically, with ping working for a while and then losing the network.
I thought maybe the NIC was faulty, but it seemed unlikely from previous test; I tried to change the cable from Server to the RJ-45 socket, something I had already tried in the preceding days; but ping was erratic again. Then remained only one possibility left: the RJ45 socket in the wall wich was new (installed from electrician a few months ago with all the others of the new lab and never used); I unplugged the cable, tried another RJ-45 and ... everything went good.
Now the server with CentOS works very good from 30 hours and 2 reboots. No more problems, the 2 NIC work perfectly.

So the electricians are to blamel. Some wire badly set and making erratic contact.
Of course I should have guessed something of the like, in fact I was too confident in the new Lab, didn't even think of it until yesterday.

But why Fedora made all that mess at reboot, while CentOS did not?
I can only guess it was testing the network lines at every reboot, found something bad in the line and for some reason I do not understand renamed the NIC in strange ways or disabled them or swapped them.

Ok now all is over, but it seems like CentOS had a more understandable behaviour than Fedora in this situation.


Paolo Bonavoglia.

[stevea]

Originally Posted by pbonavoglia

Well now the problem is over.

Not really. An incompetent admin has a new problem every day.

[QUOTE} Oooh I guess he was appliying the good old rule: "If You don't understand the problem of the client ... blame the client!"[/QUOTE]

No - I was applying the rule - if you are asked for specific information, and don't supply it, you cannot expect an answer.

Let me recap this thread:

You had SOME problem that you were incompetent to explain.
You refused to follow directions and supply needed debug information.,
Some friend on-site understood the background that you refused to supply and fixed your immediate problem.
You now live in a fantasy world where F9 has problems and Centos is all aces. ((these are essentially the same distro dude - get a clue )).

Please relabel this thread "incompetent admin makes confusion of 2INTERFACES and refuses to explain".
The problem is not your English, It's you lack of basic understanding of the tools you are trying to use and inability to follow directions.

But why Fedora made all that mess at reboot, while CentOS did not?

Answer is in the mirror.

[pbonavoglia]

Originally Posted by stevea

Not really. An incompetent admin has a new problem every day.

Let me recap this thread:

You had SOME problem that you were incompetent to explain.
You refused to follow directions and supply needed debug information.,
Some friend on-site understood the background that you refused to supply and fixed your immediate problem.
You now live in a fantasy world where F9 has problems and Centos is all aces. ((these are essentially the same distro dude - get a clue )).

So You must have Godlike powers, being able to see in great detail what happened here, some thousand km from there!!!!!
Really a good example of scientific method!!!!! Transoceanic guessing!!!

Unfortunately Your Godlike Guess is completely wrong here, If there is someone here who lives in a fantasy world ... it's not me!

Just a bit of logic:
If I were a Linux-guru I wouldn't have wasted my time with this Fedora forum!
And if I had such a friend who was a Linux guru ... just the same!

BTW, CentOS and Fedora are basically free Red Hat, very similar, but of course not identical. You should know that even a little difference in code can produce very different output.

It's just a fact that installing FC 9 on the server connected to the erratic RJ-45 produced a huge mess and different behaviour at every reboot.

It's just a fact that installing Centos 4.7 on the same server connected to the same erratic RJ-45 produced no problem at reboot, and a more manageable situation until I had the idea to change the Rj45 socket.

These are facts.
Maybe it was luck, maybe it was something else, but these are interpretations of the facts.

Anyway this conversation is by now completely useless and a waste of time, so I close here.

Paolo Bonavoglia

[stevea]

Originally Posted by pbonavoglia

So You must have Godlike powers, being able to see in great detail what happened here, some thousand km from there!!!!!

I can read what you wrote from a great distance ! Perhaps you consider that "Godlike".

Anyone can verify items 1/ and 2/ from your posts.
You stated item 3/ yourself - I have no reason to think you would lie.
Item 4/ is educated conjecture from years of dealing w/ Linux and people who cannot follow or understand directions. 98% probability that it's your config at fault You just haven't botched up your Centos YET !

Have a nice day. But sincerely - go back and read what I initially asked for. You NEVER provided the info did you ? You repeated the same useless info didn't you ? If you want help you have to cooperate within reason..

[marcrblevins]

Welcome to Fedora or CentOS Paolo!

You need to get those ethernet plug tester and find all the bad ones to be redone or you will have more headaches.

Shouldn't your router tell you you got bad cable?

Give Fedora another try when time permits.

[pbonavoglia]

Originally Posted by stevea

I can read what you wrote from a great distance ! Perhaps you consider that "Godlike".

Anyone can verify items 1/ and 2/ from your posts.
You stated item 3/ yourself - I have no reason to think you would lie.
Item 4/ is educated conjecture from years of dealing w/ Linux and people who cannot follow or understand directions. 98% probability that it's your config at fault You just haven't botched up your Centos YET !

Have a nice day. But sincerely - go back and read what I initially asked for. You NEVER provided the info did you ? You repeated the same useless info didn't you ? If you want help you have to cooperate within reason..

Mmmm this thread is becoming really useless, but quite funny too:

You say You founded Your Godllike guesses on what I wrote. Unfortunately it's definitely not so, and anyone who had the time to read this thread can verify this.

You say I gave useless info; "useless" is a relative term, surely it was useless to You; Read it again and it was a lot of information. I must concede it was founded on a misleading idea. My real mistake was to think, after having checked the NIC and the cables, it was a software problem. As a matter of fact, and as I wrote, it wasn't. Just a RJ-45 socket with an erratic contact, Since I plugged the cable in another socket all problems with the network were solved..

Then You kindly hint I have a problem every day ... I wrote I had installed many times Fedora without problems; above all I installed FC1 on a HP server in 2004, with web server, php, mysql, dns server and had almost no problem for 4 years. This is why FC9 was my first choice for the new server. Sorry, but Your guess (or maybe wishful thinking) is completely wrong.

You guessed some friend of mine solved this problem for me; this is again Your wishful thinking, I never wrote something of the like; I wrote I just unplugged the cable from a RJ-45 and plugged in another. I wish I had such a guru for friend!

An yet You now guess I made some config mistake with FC9. Again I must remember what I wrote and You still refuse to read: it was just a hardware (RJ-45) problem; all the rest is just Your dream or wishful thinking, not the slightest relation with what I wrote Anyone who had the time to read this thread can see it.

And You told FC9 and CentOS are in fact identical; yesterday I tried two GD libs (gd.so) for PhP, one was labeled FC9 the other for CentOS: the FC9 gd.so didn't work on CentOS, while the second worked perfectly. So FC9 and CentOS are not quite identical. even the PhP libraries are different, in fact they are labeled differently.

It's likely FC9 would have worked fine if I had used a good RJ45 socket from the beginning; anyway CentOS in now working good, so I have no reason to change again, and i have to give my farewell to this forum.

Maybe some times (98% ?) Your guesses were right, but this time (2% ?) they were all wrong.

Maybe You should be more cautious in Your conjectures, and not so boldly confident of Your past experiences ... it can mislead. And remember: between the White (a Linux guru) and the Black (a Linux newbie) there are almost infinite Grays ...

Have a nice day!

Paolo Bonavoglia

[pbonavoglia]

Originally Posted by marcrblevins

Welcome to Fedora or CentOS Paolo!

You need to get those ethernet plug tester and find all the bad ones to be redone or you will have more headaches.

Shouldn't your router tell you you got bad cable?

Give Fedora another try when time permits.

Yes this is a good advice.

Our lab is a school lab with very limited budget, and just two people, me and a technical aid, working here.
In the past I have used my laptop to check sockets, but of course an ethernet tester would give more information.

The job to check sockets in the new labs was up to the electician firm who cabled them; I now suspect they didn't checked all the sockets as they should have done. In fact this is the first time we find a bad socket since the lab was cabled, but there are still a few socked never used here. The sockets are connected to the LAN by a switch, not a router. I didn't notice flashing lights however.

So maybe it's now a good idea to buy a ethernet tester for our lab.

Paolo Bonavoglia.

[David Becker]

Originally Posted by pbonavoglia

...

So the electricians are to blamel. Some wire badly set and making erratic contact.

How were those cables wired/crimped?

Like one of these setups?:

http://en.wikipedia.org/wiki/TIA-568B#Wiring

Or did the electricians just match cable colors to pin number on both ends of the cable?

David