Figure 5. Basic HPSS Metadata & Filesystem Allocation
On the left hand side of the diagram, the raw physical volumes are shown attached to the disk array controller. The configuration of the disks by the controller and its software should be divided into three separate LUNs: 1) HPSS Filesystems and DB2 Backups, 2) DB2 Logs, 3) and the DB2 Tables. One disk may be kept as as a "hot spare" in the event that one of the other disks fails. This allows the system to automatically replace the failed media and rebuild the data without immediate intervention from an operator or administrator. Recommended configurations for the LUNs is to use RAID 5, which is good for small, random data access, for the first and third LUN. For the LUN associated with DB2 Logs, it is suggested that RAID 1 (Mirroring) be used since these logs are typically very large and accessed sequentially by the system. Once defined, these LUNs are assigned to the core server host via the LVM (AIX) into a Volume Group or by the operating system (Linux) into a hard drive identifier. The last step is to allocate individual filesystems and logical volumes(AIX)/partitions(Linux) as defined on the
HPSS requires the use of DB2 log mirroring. This protects the active DB2 log from possible loss due to the primary metadata disk array failure or other unavailability. To safe guard the DB2 log, separate disk/controller components are required for the LUNs providing storage for the filesystem where the log is written as well as where the archived copies of the logs are stored before being backed up to tape media. It is imperative that the components be totally separate from the primary metadata disk array (i.e. different disks, different controllers, different path such as the FC connections) to provide protection against failures. Specific resource assignment for the the
HPSS Installation Guide | July 2008 |
Release 6.2 (Revision 2.0) | 70 |
