M2200/HSG80 error in Windows 2000

[Karen Elsbree]

Win2K machine via a switch to M2200 w/HSG80 disk enclosure. I am receiving the following error in Windows.

Event ID: 51
An error was detected on device \Device\Harddisk4\DR4 during a paging operation

Disks don't appear to have any problems....

[Eugeny Brychkov]

Do you see ID9 events (timeouts)? Check HBA drivers, FC cables, switch diagnostic output
Eugeny

[Karen Elsbree]

No ID 9 errors...

[Eugeny Brychkov]

Did you set disk array up to work in w2k mode? See http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xe1a93a1c04ffd61190050090279cd0f9,00.html
Eugeny

[Andrew Cowl_1]

Here's a little note that may help.

According to HP this customer advisory is now a blanket CA, not just for clustered systems, and should be applied to all Microsoft servers with the specified drivers.


DESCRIPTION All Windows SAN installations require that the HBA driver parameter, ResetTPRLO be set to a value of 1. Specifically, when the KGPSA driver v4.53a7 or a later driver is installed this parameter needs to be changed in the registry from the default value, which is set to 0. This applies to: Compaq PCI FC Host Bus Adapter DS-KGPSA-BC, HBA driver V4-4.53a7 (NT) or V5-4.53a7 (W2K), Firmware Version 3.20x7 Compaq PCI FC Host Bus Adapter DS-KGPSA-CB, HBA driver V4-4.53a7 (NT) or V5-4.53a7 (W2K), Firmware Version 3.82a1/BIOS Version 1.60a4
SCOPE All Windows 2000 and NT4.0 clustered and standalone SAN installations, with the v4.53A7 KGPSA driver installed, require changing the ResetTPRLO parameter in the Registry to a 1. ResetTPRLO - (Third Party Logout Extended Link Service) must be added to the registry string for the CPQKGPSA entry (Windows 2000) and lp6nds35 entry (Windows NT 4.0) and set to a value of one (1). This setting forces the node reservations on LUNs to be freed in the event of a cluster failover. If this parameter is not set, a cluster failover or node crash may render the LUNs unavailable to the surviving node in the cluster. When the unavailable LUNs are accessed through the HSG80 CLI, a hung CLI may result, requiring the Restart of the HSG60/80s and the Servers in the SAN. The KGPSA-BC and DS-KGPSA-CB Readme, at the time of the writing of this CA, erroneously states the ResetTPRLO default is 1. The default is actually 0 as presented in LputilNT. One (1) was the default on earlier version drivers used in both clustered and standalone installations. This registry modification should not be required in Compaq released driver versions prior to V5-4.53a7 and V4-4.53a7. NOTE: This setting is required for both clustered and standalone Windows systems in order to clear reservations.
RESOLUTION For all Windows configurations, verify that the HBA driver parameter, ResetTPRLO is set to 1 by using the LPUTILNT utility. Launch LPUTILNT and select the first HBA. Then select Driver Parameters from the Category pane, select the Parameters button, scroll down to ResetTPRLO and double click on it. Set the value to 1, and select Permanent and Global. Click the Change button. You will need to do this for each HBA in the system. A Windows restart is NOT required after the registry modification when this method is used. This parameter is dynamic.
COMPAQ PLATFORMS AFFECTED: Non-Product Specific
OPERATING SYSTEMS AFFECTED: Microsoft Windows 2000, Microsoft Windows NT 4.0

Best of Luck.

Andy

[Karen Elsbree]

The ResetTPRLO was set according to the customer advisory several months ago.

[dentistcorn]

To ensure you have all the correct settings for your KGPSA HBA install the Multi-Bus software from ACS support software kit 8.7.

Take all the defaults when walking through the install wizard as this provides the HBAs with ALL with the appropriate settings. If you only set the "resetTPRLO" setting and not the other settings correctly your HBA will definitely hose your LUNS.This assuming your have the latest driver and firmware which is also included in the kit.

Also, be absolutely sure that your fibre pulls are disconnected from the HBA when doing performing any kind of driver or firmware upgrade. We have seen issues were several LUNS were wiped out because of this so I'm just passing along the FYI.

The order should be: 1- firmware upgrade 2- Driver upgrade 3 Multi-Bus FC switch software.