Avoiding Disk Problems

I l @ ve RuBoard

The previous section discussed how to use performance data to fix disk performance problems. Proper configuration can help you to avoid performance problems. Commands such as newfs and tunefs can be used for the High-Performance File System (HFS) to set block size and reduce the number of inodes.

The best way to avoid disk faults is to configure for redundancy and monitor critical disk resources proactively. Products exist at the hardware level that provide both data redundancy and redundant hardware components. Monitoring for hardware faults provides failure notification of redundant components . Although these faults may not cause data to be unavailable, the problems should be fixed immediately. Otherwise, a subsequent failure could cause a loss of data or availability.

Tools such as HP Predictive Support help to indicate when disks are likely to fail. You can also monitor hardware device events, so that you can replace a disk during nonproduction hours (planned downtime), instead of waiting for it to fail. Similarly, use the EMS DVM to detect the loss of a mirror, and then replace the disk before another drive fails.

At the resource level, you need to ensure that the filesystems are accessible. Monitoring the configuration and resource status can help to ensure that the filesystems are up and accessible to those who require them. Monitoring filesystem usage is critical. If a filesystem fills up, applications may have failures. By using tools such as Online JFS, you can increase filesystem capacity online, without affecting availability. By using three-way mirroring, you can perform backups on one mirror, while still maintaining data redundancy.

You should monitor free space on all disk drives to see whether you are nearing configured limits. The EMS File System Monitor can be used for this purpose. Archive files that have not been used for a long time, and remove them from the system. You can use commands such as du to view the usage for each directory and locate large files.

The availability of tools may depend on the filesystem being used. A journalled filesystem can track updates to a file, so that if the file is damaged, it can then be recovered. Online JFS is an optional product that can be used in conjunction with a journalled filesystem. Online JFS can create large file extents to avoid fragmentation, perform online filesystem resizing when a filesystem nears its configured limit, and provide data on fragmentation that is occurring. Online JFS has a defragmentation capability as well. Directories and extents can be reorganized by using the fsadm command.

I l @ ve RuBoard