Currently all our data segment files for datasets reside on the same file system. I would like to spread the segment files across a number of file systems to reduce the I/O bottleneck but have the file descriptor remain on the current file system.
What is the best method of achieving this?
Moving 'dataset' data segment files to new file systems
Moderators: chulett, rschirm, roy
Copy the datasets using either the orchadmin command or a parallel DS job, using a configuration file which specifies the disk resources you wish to write the data segment files to.
Another possible option from the hardware perspective would be to replace the existing storage with a well-distributed RAID system mounted to the same location. Backup your existing filesystem, replace the storage and then restore the filesystem onto the new storage.
You cannot simply move the segment files to a new location.
Regards,
Another possible option from the hardware perspective would be to replace the existing storage with a well-distributed RAID system mounted to the same location. Backup your existing filesystem, replace the storage and then restore the filesystem onto the new storage.
You cannot simply move the segment files to a new location.
Regards,
- james wiles
All generalizations are false, including this one - Mark Twain.
All generalizations are false, including this one - Mark Twain.
No. If you need to preserve the filename of the descriptor, rename it, then perform the copy into the original name and then remove/delete (using orchadmin) the old dataset.
Alternatively, you can always recreate the datasets using the jobs which created them in the first place, overwriting the existing name.
Regards,
Alternatively, you can always recreate the datasets using the jobs which created them in the first place, overwriting the existing name.
Regards,
- james wiles
All generalizations are false, including this one - Mark Twain.
All generalizations are false, including this one - Mark Twain.