This discussion is archived
10 Replies Latest reply: Aug 2, 2012 7:56 AM by Dan.Norris RSS

Disk Group DBFS

SPA2 Newbie
Currently Being Moderated
Hi all,

Oracle created via there installation a disk group DBFS which is really large 800GB.
Actually we don't need this space but we would like to use it for other tasks.
So it seems to be two possibilities:
- Create a Filesystem via ACFS.
Could someone describe the tasks?
- Deallocate the disk group DBFS.
Has somebody done these kind of job and could describe it?

thanks and regards

SPA2
  • 1. Re: Disk Group DBFS
    jjmcgov Newbie
    Currently Being Moderated
    How did you confirm 800gb is allocated?
    Did customer use DBFS before with LOBs for example? Deleted the LOBs and now DBFS won't shrink?
    The 800gb is the size without ever using DBFS?

    Can you confirm size with DBMS_SPACE?
  • 2. Re: Disk Group DBFS
    Dan.Norris Employee ACE
    Currently Being Moderated
    ACFS isn't compatible with Exadata storage. Instead of dropping the DBFS_DG diskgroup, you should consider creating a DBFS filesystem in it. MOS 1054431.1 may help with setting that up.
  • 3. Re: Disk Group DBFS
    SPA2 Newbie
    Currently Being Moderated
    Hi,
    I check this via asmcmd and take a look to the disk groups.

    regards
  • 4. Re: Disk Group DBFS
    SPA2 Newbie
    Currently Being Moderated
    Hi Dan,
    many thanks for this information.

    We will check to reduce the DBFS disk group to get freespace back for the other diskgroups DATA, RECO.
    But to do these tasks in a highly productive system is very complicated because we don't have a system to test these tasks.

    Actually we don't understand why the ACS Support doesn't create directy a DBFS in this diskgroup.
    It makes sense while it is one of the major ways to transfer data from and to the Exadata System.

    kind regards
    SPA2
  • 5. Re: Disk Group DBFS
    robinsc Explorer
    Currently Being Moderated
    Hi you cannot change this 800Gb DBFS partition it is used to store the ocr and vote information apart from otehrs. Also since the first 2 disks on the storage servers have around 30Gb set aside for the OS the same amount has been set aside from the other 10 disks to ensure that the disk groups are symmetrical.
    So do not play with this disk group or you will end up reinstalling your ASM/CRS etc...
  • 6. Re: Disk Group DBFS
    SPA2 Newbie
    Currently Being Moderated
    Hi,
    yes you are right.
    The only chance to change the size is during the setup of the system.
    May be if I know that I need only 400GB than it makes sense to change the size.

    regards
  • 7. Re: Disk Group DBFS
    robinsc Explorer
    Currently Being Moderated
    You cannot change the size of this partition. For example I can guess that you have a quarter rack system

    as each portion of disk that is added to the dbfs diskgroup ( which used to be called the system dg) is 27 Gb in size you have
    27*10 * 3=approx 810 Gb in your DBFS .

    You cannot add these volumes to DATA or RECO without either having widely varying candidate disks in your disk group which is a major no-no. This is one of the facts of life of the exadata architecture.
  • 8. Re: Disk Group DBFS
    Ken Simpson Newbie
    Currently Being Moderated
    Dan, my understanding is that best practices for DBFS on Exadata is for use as a staging area for loading data. Any thoughts on using it instead for a pseudo cluster filesystem for housing scripts, logs, etc for the compute nodes?
  • 9. Re: Disk Group DBFS
    Dan.Norris Employee ACE
    Currently Being Moderated
    Ken Simpson wrote:
    Dan, my understanding is that best practices for DBFS on Exadata is for use as a staging area for loading data. Any thoughts on using it instead for a pseudo cluster filesystem for housing scripts, logs, etc for the compute nodes?
    I don't think there's been anything published that limits DBFS's usage to staging data. While that is a common use case, it isn't the only use case and you have pointed out a few other valid use cases. DBFS isn't a pseudo cluster filesystem - it is a true cluster filesystem just like NFS, ACFS, or others.

    Dan
  • 10. Re: Disk Group DBFS
    Dan.Norris Employee ACE
    Currently Being Moderated
    robinsc wrote:
    You cannot change the size of this partition. For example I can guess that you have a quarter rack system

    as each portion of disk that is added to the dbfs diskgroup ( which used to be called the system dg) is 27 Gb in size you have
    27*10 * 3=approx 810 Gb in your DBFS .

    You cannot add these volumes to DATA or RECO without either having widely varying candidate disks in your disk group which is a major no-no. This is one of the facts of life of the exadata architecture.
    We've often suggested to customers that they place datafiles in DBFS_DG. While the diskgroup name may suggest DBFS, that certainly isn't the only use. It is a diskgroup and can be used as such for any data that is suitable to the characteristics of the DBFS_DG diskgroup (redundancy, etc).

Legend

  • Correct Answers - 10 points
  • Helpful Answers - 5 points