Networking
cancel
Showing results for 
Search instead for 
Did you mean: 

network team & connection issue

SOLVED
Go to solution

network team & connection issue

Hi,

I have an AD with 2 Proliant Servers (DL380 G3 & DL360G4), both running W2K
server, both with integrated GB nics. Both are teamed using HP Network Team.

Problem is following:
There is a brother printer (MFC8220) that I can ping from allover the
network except from the servers. If I plug out a network cable from a
server, I can ping again. If 2 cables are connected, and work as team, I get
"Request timed out" with I ping the device.

Anyone ever had this?

Frank
11 REPLIES
Ivan Ferreira
Honored Contributor

Re: network team & connection issue

What is the team mode that are you using? Are these network cards connected to the same switch or differents switches?
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?
Robert S. Carr
Trusted Contributor

Re: network team & connection issue

Frank, from what you wrote it seems that the only device you cannot ping is the Brother printer when using Teaming. Is this correct? You can ping other PCs, printers, servers, routers, etc but not the Brother, right?

If this is case I might try using the following to see if this will fix the problem:

Bob:

NIC Teaming seems to be a frequent problem for a lot of admins and if you end up abandoning it, except for failover, don't feel so bad. For WinOSes it seems to be more problematic.

That being said you can try a few ideas to see if this will get Teaming to work properly.

Teaming is very sensitive regarding DNS data. Two adapters must use the same DNS data. If out of sync Teaming can experience problems. To sync the adaptors

i) Disable teaming. Disable the second adaptor.

ii) At the command prompt

ipconfig 0 /registerdns

iii) Now enable the second adaptor and at the cmd prompt

ipconfig 1 /registerdns

iv) Re-enable teaming. See if this helps.

Using /registerdns will initiates manual dynamic registration for the DNS names and IP addresses that are configured at a computer.

see http://www.microsoft.com/resources/documentation/windows/xp/all/proddocs/en-us/ip
config.mspx?mfr=true

Two other tips to share with teaming are:

1) Used a fixed IP address, not a dynamically assigned one.

2) Used a fixed NIC speed. Do not use Auto Negotiate

1) is seen a lot with others who have experienced teaming problems. 2) is a personal preference but it has helped me cut down on the number of variables to consider when fixing network problems.

Good luck. Report back what worked and what didn't.

Rob

Re: network team & connection issue

First of all, I have to add following information:
I used as team mode: TLB with fault tolerance.
The different nics are connected to the same hubs, DLINK Gigabit switch, both connected to GBit ports.

Second, thanks for the answer Rob, I haven't tried it yet, but I will sure do today !

Regards;
Frank
Connery
Trusted Contributor
Solution

Re: network team & connection issue

Hi Frank,
I'm an engineer on the team responsible for HP NIC Teaming development.

The most likely culprit is that the Brother printer's TCP/IP stack does not like receiving a frame that has been load balancing from a TLB Team's secondary adapter port.

Background:
A NIC Team is known by a single MAC address to clients in the network. This MAC address is only used as the source MAC address by the Primary adapter in the Team. Secondary adapters in the team use their own MAC address as the source address. This MUST be done by NIC Teaming to be compliant with IEEE standards.

For TLB, when a client ARP's for the Team's MAC address using the Team's IP address, the Team will respond with the Team's MAC address (used by the Primary adapter). So, all traffic is received by the Primary adapter. Traffic transmitted out of the Primary adapter has a source address of the Team. Traffic transmitted out of the Secondary adapters use their own MAC address (not the same as the Team MAC address). Since clients with a proper TCP/IP stack implementation should ONLY update their ARP table based on data received in ARP request or replies, receiving data with a different source MAC from a secondary adapter will not affect clients with a complient TCP/IP stack.

Some implementations of the TCP/IP stack in clients are picky in that they require the source MAC address to match the MAC address in the ARP cache. There's no good reason for this. In fact, we've found a few of our own products that had this incorrect implementation and we had to fix it.

Ways to verify the problem:
To verify this behavior, you can do a couple of things. 1. unplug all but one port in the team. If the problem goes away, that probably means this was the problem. I think you have already tried this. 2. change the team type to NFT. NFT does not load balance traffic out of the secondary ports. If the problem goes away, that probably means this was the problem. 3. If the Brother's IP address ends in an odd numbered IP address, change it to an even numbered IP address - or vice versa. The load balancing algorithm uses the last couple of bits of the IP address (if load balancing algorithm is not set for MAC based) to load balance traffic. Change from odd to even or vice versa should move the Brother's traffic from the secondary port to the primary. If the printer starts to work, then you have verified this problem.

Solutions if you verify this is the problem:
1. Use NFT instead of TLB.
2. Connect the team to a switch (not a hub) capable of doing port trunking and use SLB. This team type works with the switch and uses the same MAC on all ports.
3. Buy an Intelligent Networking Pack (INP) license and use the "Dual Channel" team type with only two ports. Put one port in Group A and one port in Group B. Each group sends it's own ARP requests/replies and acts like two independent teams under the same IP address.

Here is a link to a whitepaper about NIC Teaming. It will provide more information about INP if you are interested in it.
ftp://ftp.compaq.com/pub/products/servers/networking/TeamingWP.pdf

There is one other possibility for the cause of your problem. In NIC Teaming driver versions prior to 8.20, NIC Teaming load balanced ARP replies out of secondary ports. This caused a similar problem to the one I described above, but is slightly different. For this problem, NIC Teaming can help by forcing all ARP requests/replies out of the Primary port. So, as of 8.20 we have done this. I recommend updating to the latest version of the teaming driver to make sure this isn't your problem.
http://h18004.www1.hp.com/support/files/networking/us/download/23963.html

Hope this helps!
-sean

Re: network team & connection issue

that was an extremely usefull help. Thanks a lot!

And yes, indeed like you said, I tried plugging out all the cables except one and then I could ping the brother printer. Currently that printer "seems" to be the only problem and so I mapped it locally on the clients. But that is not the way to work of course.

I will definately test all those remarks as soon as possible !

Thanks again!
Lucky Luciano
Regular Advisor

Re: network team & connection issue

Hi

I would add one observation and it is to connect to switch console and to observe behaviour of ports included in team.
I am not saying that this will solve problem with printer, but could be interresting to see if there is a different approach
I admire your fashionable running shoes!

Re: network team & connection issue

Sorry Peter,
cannot quite follow that part, can you explain bit more?
Server is connected to iLo aswell, so I can monitor behaviour even from remote. (Because I can only go there again next week Tuesday)

Frank
Robert S. Carr
Trusted Contributor

Re: network team & connection issue

Frank:

Be hard to add any more info to Sean's excellent reply but just FYI:

Trunking is a term I have seen with HP Procurve switches. Whether the features are indentical I don't know but Cisco (we use Cisco here) uses a feature called Fast EtherChannel which does basically the same thing. It allows the same MAC or IP address to be bound to two or four Ethernet adapters, providing parallel links.

As Sean mentioned you can use Switch-Assisted Load Balancing (SLB) which has the benefit of load balancing for both incoming and outgoing traffic. The caveat is the switch MUST be able to support this feature.

Rob

Re: network team & connection issue

Sean, Robert, you guys are geniuses! Thanks, that Switched assistent load balancing did it alright!
Sean's explanation was perfect, Robert made a little addendum that made it all clear, crystal clear.

Thanks guys !

Frank
Douglas G. Arnett
Occasional Visitor

Re: network team & connection issue

Ok - I read through this thread and it seems to address my need (maximum bandwidth for a server), but I have one simple question (sorry, a bit of a n00b one):

Once I have the HP team setup, and set to SLB, do I need to change the switch configuration at all?
Connery
Trusted Contributor

Re: network team & connection issue

Douglas,
Yes, you need to configure the switch. If it's a Cisco switch, you need to configure the associated switch ports for EtherChannel.

See Teaming whitepaper for details:
ftp://ftp.compaq.com/pub/products/servers/networking/TeamingWP.pdf

Best regards,
-sean