Moving a 20TB Database from ASM to NFS

One of the recent challenges I had was to migrate a 20TB database from ASM to NFS and then decommission ASM. This post will give an overview of how I achieved it in relation to the requirements.

Currently our storage of choice for infrastructure is NFS on Netapp and most of our existing databases are already on there. Our Data Warehouse platform is a 2 node RAC running on Red Hat 5.5 with VMAX storage.

A weekend outage window was available but I wanted to keep downtime to a minimum as there was a lot of testing and other factors in play. So my basic migration method was as follows:

  • RMAN “backup as copy” datafiles to NFS mounts
  • Switch datafiles and recover
  • Move votedisk and ocr to NFS
  • Remove any ACFS
  • Remove any ASM dependancy
  • Stop and disable ASM
  • I planned for a 2-4 hour outage window. The “backup as copy” can run while the database is online and the outage would start when I needed to switch and recover the datafile copies until ASM was removed.

    Our database standard is to layout datafiles across various mounts so my RMAN “backup to copy” script needed to be broken up into datafile copies and I couldn’t just do a “backup as copy database format…”. I generated my rman script with:

    So my rman script looked something like the following. I set parallel of 8 so needed 8 datafiles per “backup as copy” to have the desired throughput.

    I also generated a similar switch and recover script with the following:

    And that looked something like this:

    I moved temp and undo a few days before the migration during the quieter periods so I was already to go prior to the weekend run. Testing this in other environments and understanding the infrastructure we knew we’d get around 1TB an hour for our datafile copies so our backup to copy job was going to run for around 20 hours.

    The “backup to copy” job was kicked off on a Friday afternoon and I was ready to do the recovery and remove ASM starting at noon on Saturday. Nightly batch processing was stopped to minimise the amount of work done during the recovery.

    The datafile switch and recovery was done in under an hour. I then proceeded to move the votedisk and ocr off ASM.

    So once I’d moved the vote disk and ocr all that was left was to remove acfs, stop a few diskgroups and disable asm.

    The outage was about 2 hours and I had everything migrated from ASM to NFS and ASM was no longer running or enabled.

    Leave a Reply

    Your email address will not be published. Required fields are marked *

    CAPTCHA * Time limit is exhausted. Please reload CAPTCHA.