10.07.2015 Views

POWER SOLUTIONS

POWER SOLUTIONS

POWER SOLUTIONS

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

STORAGE TECHNOLOGYmemory manager was attempting to copy data to or from memoryand experienced a problem. For example, the following could appearin the Microsoft SEL:Event ID: 51Event Type: WarningEvent Source: DiskDescription: An error was detected on device \device\harddisk0\DR0 during a paging operationEvent ID: 11Source: %HBA_Driver_Name%Description: The device, \device\ScsiPort0, did not respondwithin the time-out periodAnother indicator is a blue screen, which signifies an OS crash.If administrators see the following messages, then latency issuesmay be the culprit:0x00000050 PAGE_FAULT_IN_NONPAGED_AREAor0x0000000A IRQL_NOT_LESS_OR_EQUALA quick resolution to the preceding types of problems is to placea disk within the server for page and swap files. This approach canhelp prevent access from being influenced by devices, fabrics, orhosts on the SAN.Linux dependenciesLinux has many of the same dependencies as Microsoft Windowswith regard to host segregation, LUN 0, and exclusive accessto the bootable disk. However, Linux does not have the samelatency sensitivities associated with page and swap files asMicrosoft Windows because the Linux kernel is not a pageableOS—that is, it does not swap for kernel operations, but it willswap for applications. Linux pages applications in and out afterthe boot sequence. If latency occurs when accessing the LUN(more than 60 seconds without response), SCSI time-out messagessimilar to the following will appear on the console andpossibly in the system log:kernel: scsi : aborting command due to timeout : pid50212683, scsi0, channel 0, id 0, lun 0 Read (10) 00 0048 01 8f 00 00 80 00kernel: scsi: device set offline - command error recoverfailed: host 0, channel 0, id 0, lun 0kernel: SCSI disk error: host 0, channel 0, id 0, lun 0,return code = 6000000kernel: I/O error: dev 08:11, sector 22903784If the system is configured with multipathing HBA drivers, afailover should occur before a kernel panic is initiated. Linux doesnot automatically mount file systems on all discovered LUNs asWindows does. By default, the Linux OS accesses only the LUNsspecified in the /etc/fstab file. Most Linux file systems are notcluster aware, so administrators should not mount the same LUNand file system on two servers simultaneously because file corruptionmay occur.A properly designed SAN can help minimize the possibility ofa kernel panic. Data center evaluations and SAN consulting candramatically reduce the current risks associated with boot-from-SANtime-out issues.Recovering from failure in a boot-from-SAN environmentBooting from a SAN is a powerful capability, and in a disaster recoveryscenario, it is critical. Thus, administrators must know how toreplace a mission-critical server in a boot-from-SAN environment.To perform this task, administrators should adhere to the followingbest practices:• When replacing a server, administrators should remove theHBA from the failed server, install it in the replacementserver, and configure the system BIOS in the replacementserver to boot from the Fibre Channel HBA.• If the replacement server is a different model from the one itis replacing, either the OS will reconfigure itself and promptthe administrator to insert the driver CD for the replacementhardware, or the administrator will have to manually installthe driver for the replacement hardware. The OS will notrequire the reconfiguration of applications such as MicrosoftExchange Server.• If an HBA is replaced by an HBA of the same model or fromthe same vendor, administrators should update the accessrights on the RAID ports to reflect the change in the WorldWide Name (WWN) of the replacement HBA. The WWN ofthe Fibre Channel HBA can be found by accessing the HBA’sROM extension utility during system power-up. In addition,administrators must configure the HBA to boot the serverfrom the appropriate LUN.• If the replacement HBA is from a different vendor, eitherthe OS will reconfigure itself and prompt the administratorto insert the driver CD for the replacement hardware, or theadministrator will have to manually install the driver for thereplacement HBA.• For x86-based systems, if a RAID controller fails duringthe reboot of a server containing two HBAs, the designatedboot LUN may not be visible on the HBA’s primarypath. In this case and depending on the system BIOSdesign, the system may boot from the path configured86DELL <strong>POWER</strong> <strong>SOLUTIONS</strong> Reprinted from Dell Power Solutions, August 2005. Copyright © 2005 Dell Inc. All rights reserved. August 2005

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!