proliant DL380 G3 w/ NC7782 experienceing packet loss

[Mark Nelson_2]

We just moved 2 of the DL380 G3's into a gigabyte networking space and we are now getting between 3% and 10% packet loss between those two servers.

As a control, I shutdown the interface in the new address space and brought up the other nic that is still in the old 100m/b network with 0% packet loss.

Placing a sniffer on the connection between the two servers shows that server 1 sends a frame to server 2 and server 2 does not responed to that frame sometimes for upto 2 seconds.

I've updated the NC7782 network drivers on both servers but this has not had any affect.

The switch is not showing any errors as well.

What else should we do/look at?

[Tarek Kawach]

Hi Mark:

What is your current driver version for the nic card and the network teaming config. utilities.

What is your setting for the nic card and are they teamed
example: 100/Full or set to auto auto.

What is your switch setting and what kind of switch to your have

Since this issue occure; when you moved the server to gig switch, we need to make sure that the switch and nic card are set to the same speed and duplex.

- the new HP network and teaming configuration utility has some diagnostics you can run to test the nic card.

- if you set your switch to Auto/Auto and your card to Auto/auto. do you have any dropped packetts.

- What is the fram size you are using (example Jumbo frame)

-- There is lots of variables to this issue, I don't think it is hardware, we need to get a cross over cable and try the nics with other machines or workstation to see if you still get dropped packets ; possible the switch port or switch settings.


- Here is some links for you that might help you.

Main Link: HP ProLiant DL380 G3, Microsoft Windows Server 2003 http://h18000.www1.hp.com/support/files/server/us/locate/69_4706.html#14
-
HP NC10xx/67xx/77xx/150x/320x/324x/325x/326x Gigabit Ethernet NIC Driver for Windows Server 2003
version 8.48.0.0 (14 Nov 05)
http://h18000.www1.hp.com/support/files/server/us/download/23652.html
-
HP Network Configuration Utility for Windows Server 2003
version 8.15.0.0 (10 Nov 05)
http://h18000.www1.hp.com/support/files/server/us/download/23442.html
-
HP ProLiant Fast Ethernet & Gigabit NIC Diagnostics and Documentation
version 7.51 A (4 Feb 04)http://h18000.www1.hp.com/support/files/server/us/download/20199.html
-

- another thing we can try to do ; is to uninstall and reinstall TCP/ IP protocal; that will reset your TCP/IP Stack and registry.

- But my thinking here is the only thing that was changed is the Switch.

Make sure everything is updated on the server
- Software and Firmware.

- here is the firmware CD - that will bring your server to the latest bios updates etc.

- possible we got resource issues in the bios.
Firmware Maintenance CD version 7.40 (9 Sep 05) http://h18000.www1.hp.com/support/files/server/us/download/23331.html
-
ProLiant Support Pack for Microsoft Windows Server 2003 version 7.40 A (9 Sep 05) http://h18000.www1.hp.com/support/files/server/us/download/22871.html

PS: Run latest Support pack and firmware update on the server; the firmware cd is bootable and will do all your firmware for the server.

Best regards,

Tarek.

[Mark Nelson_2]

first, terak, thank you for such a comprehensive reply!
**
What is your current driver version for the nic card and the network teaming config. utilities
**
The driver date is 10/17/2005
The Version is 8.48.0

**
What is your switch setting and what kind of switch to your have
**
I can't tell you the specific make and model of switch.

I can tell you that we have forced it down to 100/full with the same packet loss results as auto on both the switch and servers

**
Since this issue occure; when you moved the server to gig switch, we need to make sure that the switch and nic card are set to the same speed and duplex.
**
We did test to make sure that the switch and servers were both forced to the same speeds of 100/full and auto. The NC drivers don't appear to have a 1000/full setting. However when we set the servers to auto, and the switch to auto, it reported the servers connecting at 1000/full.

In either case, with it forced to 100/full or auto, packet loss occured.

**
- the new HP network and teaming configuration utility has some diagnostics you can run to test the nic card.
**

I apologize for being ignorant here, but what new networking and teaming utilities?


**
- if you set your switch to Auto/Auto and your card to Auto/auto. do you have any dropped packetts.
**

Yes

**
- What is the fram size you are using (example Jumbo frame)
**

Again, I apologize, but I don't know how to tell.

**
-- There is lots of variables to this issue, I don't think it is hardware, we need to get a cross over cable and try the nics with other machines or workstation to see if you still get dropped packets ; possible the switch port or switch settings.
**

I'll try to convice the networking folks of this. Unfortunately I'm in california and the servers are in Dallas Texas.

**
links and firmware..
**

I will certianly read the links and see what we can do!

Thank you and I'll get back with a reply on the updated firmware/software results!

Thank you again,

Mark Nelson

[Mark Nelson_2]

UPDATE:

We are definatly not using Jumbo Frames.

Also in monitoring the 2nd severs network interface for errors via the HP Network Configuration Utility, it suddenly started throwing an error about a corrupted XML file and wants us to re-install.

Must be a Monday.

[Mark Nelson_2]

Well the driver pack is installed with no change. Sever packet loss still on the gig network with zero errors on the switch or nic's.


Frame sizes are not jumbo. We are completely befuddled here and continue to search for solutions.

We are hoping to get the firmware packs installed tonight.

Any other suggestions on this?

[Mark Nelson_2]

Well, we have updated the firmware and drivers and anytime we have the nics connected to the gigabyte network turned on, we get packet loss of between 3 to 5% (once as high as 10%).


We are at a complete loss now.

[Tarek Kawach]

Did you try one nic at a time at full/duplex 1000 speed.

it is normal to get some packett loss. but no more than 5%.

- Try a cross over cable to another gig nic card on different server or workstation.

to eliminate all other variables.

- if you have 3rd party nic card like 3com or intel - that will eliminate the drive corruptions in windows.

- Let me know what happens if you try only one nic at a time with Cross-over cable straight to other nic at full speed.

- is this the only machine or server that is causing the issues. you said you updated from 2000 to 2003. Possibl