解决SAP PI Cluster系统故障

文档已经交付给用户了，这里总结一下：

SAP PI的PI服务当在MSCS群集的node1和node2中都启动的时候，MSCS故障，所有PI资源组会在node1和node2中来回切换，导致Oracle OFS资源和MSCS资源也切换，由于PI占用内存很大，有30GB内存，这样的自动来回切换约8次后，pubilc网卡down，up多次崩溃。

由于MSCS切换和OFS资源切换都没有问题，检查MSCS的集群配置参数，无误。

检查操作系统，看是否有不利于MSCS的补丁，无误

检查网络设置和网卡属性中的BOE offload，RSS，speed，无误

检查针对WINDOWS 2003 R2 SP2中的伸缩端缩放，补丁已达，无误

检查public和private的网络千兆交换机环境，无误

最后发现：

node1和node2的网卡 HP NC357i驱动都是最新的556版本，而node1 的网卡固件是 527版本，node2的网卡固件是534，经查确认，527固件和556驱动不匹配。找到问题了

解决，驱动由于是最新，不必重装驱动，刷固件

C:\SWSetup\SP50817>nxflash_x64.exe -i private --all
0/8 - Init
*** Currently in flash ***
Board Type       : HP NC375i Integrated Quad Port Multifunction Gigabit Server Adapter
Firmware Version : 4.0.534
MAC Address 0    : 68:B5:99:C4:B2:B8
MAC Address 1    : 68:B5:99:C4:B2:B9
MAC Address 2    : 68:B5:99:C4:B2:BA
MAC Address 3    : 68:B5:99:C4:B2:BB
Serial Number    : 牋牋牋牋牋牋牋牋牋牋牋牋牋牋牋牋??
NIC binary romimage found in C:\SWSetup\SP50817
Rom Image        : C:\SWSetup\SP50817\phantom_romimage
1/8 - Extracting Romimage
Firmware version From Board: 4.0.534
Firmware version From Romimage: 4.0.539
WARNING: This operation will take the NIC offline.
Do you wish to upgrade? (Y/N) y
Disabling devices
Disabling devices
Disabling devices
Disabling devices
Driver Loaded in Quiesce mode
2/8 - Restoring License
 100%  - DONE
 100%  - DONE
No vNIC property area in romimage
No VPD area in romimage
3/8 - Calculating MD5
 100%  - DONE
4/8 - Backing up current flash
 100%  - DONE
Backup file : "flashbackup__v4.0.534_Sat-Oct-13-22-06-50-2012" - completed successfully.
5/8 - Updating flash
WARNING: This is a very sensitive operation.
Do not interrupt until operation is complete.
setting up the flash_write
 100%  - DONE
6/8 - Verifying Flash MD5
Flashing completed successfully.
Reboot system for firmware to take effect
Enabling devices
Enabling devices
Enabling devices
Enabling devices
Driver Loaded in Normal mode
7/8 - Performing cleanup
8/8 - Finished

C:\SWSetup\SP50817>

在2号机node2上

C:\SWSetup\SP50817>nxflash_x64.exe -i private --all
0/8 - Init
*** Currently in flash ***
Board Type       : HP NC375i Integrated Quad Port Multifunction Gigabit Server A
dapter
Firmware Version : 4.0.527
MAC Address 0    : 68:B5:99:B3:3C:58
MAC Address 1    : 68:B5:99:B3:3C:59
MAC Address 2    : 68:B5:99:B3:3C:5A
MAC Address 3    : 68:B5:99:B3:3C:5B
Serial Number    : 牋牋牋牋牋牋牋牋牋牋牋牋牋牋牋牋??
NIC binary romimage found in C:\SWSetup\SP50817
Rom Image        : C:\SWSetup\SP50817\phantom_romimage
1/8 - Extracting Romimage
Firmware version From Board: 4.0.527
Firmware version From Romimage: 4.0.539
WARNING: This operation will take the NIC offline.
Do you wish to upgrade? (Y/N) y
Disabling devices
Disabling devices
Disabling devices
Disabling devices
Driver Loaded in Quiesce mode
2/8 - Restoring License
 100%  - DONE
 100%  - DONE
No vNIC property area in romimage
No VPD area in romimage
3/8 - Calculating MD5
 100%  - DONE
4/8 - Backing up current flash
 100%  - DONE
Backup file : "flashbackup__v4.0.527_Sat-Oct-13-21-06-28-2012" - completed succe
ssfully.
5/8 - Updating flash
WARNING: This is a very sensitive operation.
Do not interrupt until operation is complete.
setting up the flash_write
 100%  - DONE
6/8 - Verifying Flash MD5
Flashing completed successfully.
Reboot system for firmware to take effect
Enabling devices
Enabling devices
Enabling devices
Enabling devices
Driver Loaded in Normal mode
7/8 - Performing cleanup
8/8 - Finished

C:\SWSetup\SP50817>

问题解决！

后来和采购确认，两台机器来源采购相差半年，不是同一批次。2号机是开发机，半年后才新购1号机生产机，然后实施的时候开发机和生产机做MSCS PI。

看来实施MSCS的人技术很毛躁，不靠谱。Windows企业环境要更加精细化，对技术素养要更高，因为很多错误你无法深入内核解决，我不可能遇到问题就看dump崩溃核心转储文件，或者拿出windbg就开工。——当然这是最后的办法