UNSOLVED

admingirl

updated

16 years ago

A

admingirl

2 Intern

614 Posts

0

5164

February 16th, 2011 14:00

adding a hyper to an existing meta in production

Hello,

Our sql cluster is running in windows 2003. I have been asked to add 80GB to an existing mount point. I asked the windows admin if there is something like veritas volume manager where I would be able just to add the meta to a disk group and expand the file system but there is not and so I need to validate what to do. Can I add a device to a meta that has already been masked, mapped? Do I have to unmask it? These are not BCV(s).

Have the windows admin offline the drive(meta)

I add the drive(s) to the current meta, will be 10 hypers

Thanks in advance!

admingirl

  • dynamox

    11 Legend

    20419 Posts

    87439 Points

    1277

    1

    Posted February 16th, 2011 14:00

    in most cases you can do it online. If current device is a concatenated meta, you simply add new members and expand partition with diskpart. If current device is a striped meta, you will need to use a BCV (2-way mirror BCV only) to expand the meta with "protect_data" parameter, then diskpart. If existing device is a regular device, it will need to be unmapped first and then expanded, then diskpart.

  • admingirl

    2 Intern

    614 Posts

    1277

    0

    Posted February 17th, 2011 06:00

    Thank you Dynamox, I will run this past our windows admin. Once we get the meta number, I will check to see what it's attributes are.

  • 1277

    0

    Posted February 17th, 2011 06:00

    Hi Dynamox,

    i have a small doubt regarding your statement "If existing device is a regular device, it will need to be unmapped first and then expanded, then diskpart." Do you mean if the device is just one device (non meta) ?

    Regards,

    Saurabh Rohilla

  • dynamox

    11 Legend

    20419 Posts

    87439 Points

    1277

    0

    Posted February 17th, 2011 06:00

    correct

  • dynamox

    11 Legend

    20419 Posts

    87439 Points

    1277

    0

    Posted February 17th, 2011 10:00

    absolutely

  • 1277

    0

    Posted February 17th, 2011 10:00

    but can we create a meta without losing the data?

    regards,

    Saurabh

  • 1277

    0

    Posted February 17th, 2011 12:00

    Converting the regular device to striped meta will result in DATA LOSS but you can convert regular to concat while preserving data.

    regards,

    Saurabh

  • 990

    0

    Posted February 17th, 2011 14:00

    In SE 7.2 u dont need to unmap while converting regular to concat but the old rule still apply for striped.

  • dynamox

    11 Legend

    20419 Posts

    87439 Points

    990

    0

    Posted February 17th, 2011 16:00

    just 7.2 or 7.2 + VMAX + 75 ?

  • Sam Claret

    86 Posts

    635

    0

    Posted February 17th, 2011 23:00

    Hello

    If you are adding a hyper to a non meta then you can set the non-meta as the meta head - add the additional hyper as a concat. Next if you want stripped you

    can convert the concatonated meta you now have to a stripped meta - using protect_data=TRUE and a BCV to protect the data.

    If on the otherhand it is a meta you are adding to then if it is stripped then you would use the protect_data=TRUE and a BCV to protect the data. If it is concatonated you do not need the protect_data parameter or BCV to maintain the data on the origonal disk.

    Disk expansion can be done with the disk online - it is the additional hyper that has to be in a unmasked state.

    All the scenarios above keeps the data on the part of the disk/meta that contains the meta-head.

    Just in case my wording isnt clear the 4 scenarios below maintain the data on the origonal device - however the data on the additional hyper is lost.

    non-meta + hyper -> concat meta (no need for protect_data=TRUE or BCV)

    concat meta + hyper -> concat meta (no need for protect_data=TRUE or BCV)

    striped meta + hyper -> stripped meta (MUST have protect_data=TRUE and BCV)

    concat meta -> convert to stripped meta (MUST have protect_data=TRUE and BCV)

    Thankyou

    Sam Claret EMC TSE3