Poor send performance on DL380G5 with NC373i

[Tom Riley1]

Christope Iasci;

Were the updates on Broadcom's site? Are you confident it resolved the problem?

[Erik Briggs]

So far we have tried every "fix" in this thread and have not resolved the issue. Running a DL380 G5, using one NIC on a Gigabit connection. All cables and switchports have been quadruple-checked. Latest Firmware on the NIC's, Latest Drivers from HP, Latest Configuration Utility. Changed the RSS and Chimney settings in the registry AND in the HP utility. Turned off TOE in the utility. Ran the netsh commandline found above, and rebooted.

Here is the portion of my robocopy log that applies:

New File 0 2007/11/07 17:09:34 cum_altreftype_t
New File 0 2007/11/07 17:03:40 cum_anetypecode_t
New File 456.2 m 2007/11/07 02:18:42 cum_apptmt_t
Newer 2527 2007/11/07 17:03:43 cum_billareacode_t
Newer 114.3 m 2007/11/07 17:08:54 cum_charge_t_1
Newer 50.1 m 2007/11/07 17:12:34 cum_charge_t_2
New File 314.2 m 2007/11/07 02:14:37 cum_claim_t
2007/11/07 13:20:04 ERROR 64 (0x00000040) Changing File Attributes F:\chronos\cum_claim_t
The specified network name is no longer available.

Waiting 5 seconds... Retrying...
New File 314.2 m 2007/11/07 02:14:37 cum_claim_t
Newer 159 2007/11/07 17:03:43 cum_claimadjgrp_t
Newer 74958 2007/11/07 17:03:43 cum_claimadjrsn_t
New File 285.1 m 2007/11/07 02:19:29 cum_claimhistory_t
2007/11/07 13:24:38 ERROR 64 (0x00000040) Changing File Attributes F:\chronos\cum_claimhistory_t
The specified network name is no longer available.

Waiting 5 seconds... Retrying...
New File 285.1 m 2007/11/07 02:19:29 cum_claimhistory_t
Newer 4.1 m 2007/11/07 17:09:02 cum_cpacct_t
Newer 2607 2007/11/07 17:03:43 cum_cpactcode_t

Any other possible fixes? My hair is going greyer by the minute.

[Anwarul Ather]

We are facing the same problem with ML370 G5 servers. server is running W2K3 R2.

We recently upgraded to ML370 G5 and recieving lots of complains about slow performance of the server.

I will try disabling RSS and TOE on the machine, but as Suzanne as said, this cannot be termed as permanant solution.

I am also following up with my local suppier and HP guys on this issue.

Desperately looking for the solution.

[Jon Ward]

If you suffer from these issues, please report them to a HP Global Solution Center. A number of TOE related issues have been resolved using ProLiant Support Pack 7.90, but others may remain. Please open a case with HP so that your exact issue becomes known and investigated.

What makes these issues difficult to pin down is that they occur in some environments and not others. This is particularily true post-PSP 7.90. As such, as more people report these issues, more data becomes available to those who are investigating these issues.

Cases can be opened online from http://www1.itrc.hp.com/service/mcm/countrySelection.do? Alternatively, call (800) HP-INVENT to talk to an Solution Center agent.

[Tom Riley1]

To John Ward,
My customer does have a case open and it has been for quite some time. 3219812001. I've searched the forums constantly trying to find a fix to this problem. However as you indicate the symptoms are somewhat varied, and the errors range from the BSOD to a total disconnect from the network, and everything in between. The problem is not limited to Proliant servers, and from my perspective there are 3 commonalities, 1st is the fault is triggered by high data transfer, second of course is MS Server 2003 Service Pack installation, the third seems to be the standard Broadcom NIC, found in Proliants and some Dell PowerEdge servers. My customer has utilized the MS workaround and for the time being is not suffering from the loss of TOE, but of course that will change. Support feels this is solely a Microsoft problem. Incidentally moving them to PSP 7.90 did not cure the problem.

[Jon Ward]

Thomas,

thank you for responding. I can appreciate the problem that you are experiencing; somehow, the problem is positioned at the junction between three different vendors; Broadcom is involved in creating the firmware and drivers used, Microsoft is introducing a new technology that seems to have some problems (according to Microsoft Technet 936594), and HP is publishing those said firmware and drivers.

If you are not satisfied with HP's response, and you think this issue that may jepordize your relationship with HP, please engage a HP manager for a second opinion. Nevertheless, I can see that in your specific case, you have already had a lot of highly experienced people working on your issue.

Just curiously, has anyone on this thread opened a ticket with Microsoft in connection to the Scalable Networking Pack and TOE? Did they have a response beyond the proposed workarounds provided in Technet 936594?

[David Borden]

We have several DL380 G5s that are brand new. We are rolling them out as DCs and SQL servers. We are running in to all kinds of problems with network issues.

So far:
-SQL connections get disconnected randomly when the NICS get stressed
-We cannot use these servers as print servers because they cause constant communication errors
-Randomy network wierdness and slowness.

the netsh command to disable chimney helped with the SQL disconnected but we are still experiencing these other issues. HP has been giving me the runaround for 2 months. This hardware is worthless to us and I am not sure what to do as we can't roll out any more of these servers in this condition.

[Jon Ward]

Was your case ever escalated? A case should not last that long without being escalated unless action plans were delayed (waiting for downtimes, change management, etc.)

[David Borden]

It was escalated to 3rd level support on Oct 8. Nobody has contacted me. I have talked to my contact at HP and he claims to be having someone get in touch with me. We are hoping to hear back soon but given earlier responses in this thread it doesn't look promising.

[Jon Ward]

3rd level would be an engineering group. The 2nd level escalation engineer would still be the actual case manager, but an engineering group would be investigating.

If you have any questions pertaining to the specifics to your case, please contact your escalation engineer.