volume shadowing question - VMS

This is a discussion on volume shadowing question - VMS ; Hello We are installing an EVA4000 in replacement of an MSA1000. I was wondering if I could use HBVS to switch my data onto the new storage, online. That is: create a new drive on the EVA4000, mount it as ...

+ Reply to Thread
Results 1 to 8 of 8

Thread: volume shadowing question

  1. volume shadowing question

    Hello

    We are installing an EVA4000 in replacement of an MSA1000.

    I was wondering if I could use HBVS to switch my data onto the new storage,
    online.

    That is: create a new drive on the EVA4000, mount it as a shadow member of
    the existing one on MSA1000.
    When the merge is complete, dismount the member that's on the MSA1000.

    A possible technical problem:
    The VMScluster is comprised of 2 Alphaservers and one IA64 server. When I
    mount /cluster the shadow member, will all nodes see it properly (different
    architecture should not be a problem on VMS) ?

    And is there something obvious that I do not see ? Is this a workable plan ?
    I remember 5-6 years ago, HBVS would not mount members of different sizes so
    that will certainly be a problem if it's still the case.

    And I hope HP can lend me HBVS licenses for a week or two... I may still
    have some for Alpha but certainly not for IA64.

    Just thinking, looking for feedback

    Thanks
    --
    Syltrem
    http://pages.infinit.net/syltrem (OpenVMS information and help, en français)



  2. Re: volume shadowing question

    Given the way your question is worded, I'll assume you aren't already
    using shadowsets. If that's the case, you will at least have to dismount
    the current disk to re-mount it as a one-member shadowset to get the
    process started.

    But then, yes, you should be able to "transfer" the data this way. Then
    after the data is transferred you'd have to dismount the shadowset and
    remount the new disk. It'll complain though if you try to re-mount the
    device as a non-shadowset.

    Considering you will have to dismount the original disk before the
    transfer and dismount the target disk after the transfer, I don't see the
    advantage of this "shadow" tranfer method. You might as well, make the
    storage you want on the EVA, then BACKUP the content from one disk to another.
    You're going to have "downtime" on that disk either way.

    You may have licenses for shadowing on the IA64 already. If you have the
    MCOE (Mission Critical Operating Environment) license, which it seems you
    do if you have clustering, then I believe that also covers volume
    shadowing. You're good to go there.


    On Thu, 13 Sep 2007, Syltrem wrote:

    > Hello
    >
    > We are installing an EVA4000 in replacement of an MSA1000.
    >
    > I was wondering if I could use HBVS to switch my data onto the new storage,
    > online.
    >
    > That is: create a new drive on the EVA4000, mount it as a shadow member of
    > the existing one on MSA1000.
    > When the merge is complete, dismount the member that's on the MSA1000.
    >
    > A possible technical problem:
    > The VMScluster is comprised of 2 Alphaservers and one IA64 server. When I
    > mount /cluster the shadow member, will all nodes see it properly (different
    > architecture should not be a problem on VMS) ?
    >
    > And is there something obvious that I do not see ? Is this a workable plan ?
    > I remember 5-6 years ago, HBVS would not mount members of different sizes so
    > that will certainly be a problem if it's still the case.
    >
    > And I hope HP can lend me HBVS licenses for a week or two... I may still
    > have some for Alpha but certainly not for IA64.
    >
    > Just thinking, looking for feedback


  3. Re: volume shadowing question


    "Ryan Moore" wrote in message
    news:Pine.LNX.4.64.0709131254450.10305@jaipur.loca l...
    > Given the way your question is worded, I'll assume you aren't already
    > using shadowsets. If that's the case, you will at least have to dismount
    > the current disk to re-mount it as a one-member shadowset to get the
    > process started.
    >
    > But then, yes, you should be able to "transfer" the data this way. Then
    > after the data is transferred you'd have to dismount the shadowset and
    > remount the new disk. It'll complain though if you try to re-mount the
    > device as a non-shadowset.
    >
    > Considering you will have to dismount the original disk before the
    > transfer and dismount the target disk after the transfer, I don't see the
    > advantage of this "shadow" tranfer method. You might as well, make the
    > storage you want on the EVA, then BACKUP the content from one disk to
    > another.
    > You're going to have "downtime" on that disk either way.
    >
    > You may have licenses for shadowing on the IA64 already. If you have the
    > MCOE (Mission Critical Operating Environment) license, which it seems you
    > do if you have clustering, then I believe that also covers volume
    > shadowing. You're good to go there.
    >
    >


    Hi

    It`s been some time since I used shadowing and I forgot that.
    Yes of course I need to mount DSAxx: /shadow=(...) so I need to dismount
    first.

    You answered my question fully, thank you !

    Back to plan A...

    Thanks !

    Syltrem



  4. Re: volume shadowing question

    Syltrem wrote:
    > The VMScluster is comprised of 2 Alphaservers and one IA64 server. When I
    > mount /cluster the shadow member, will all nodes see it properly (different
    > architecture should not be a problem on VMS) ?


    Correct.

    > there something obvious that I do not see ? Is this a workable plan ?
    > I remember 5-6 years ago, HBVS would not mount members of different sizes so
    > that will certainly be a problem if it's still the case.


    If it works, you need to also ask the question of whether your larger
    disks on the new system will be usable. Say you have 1 gig drives on the
    old one and you introduce a 10 gig drive into the shadowset. After
    you've remove the older 1 gig drives, your shadowset will still only
    have a 1 gig capacity because that is how the drive was initialised when
    it was included into the original shadowset.

    Not sure if I read it or dreamt it, but doesn't VMS now have something
    like dynamic volume resizing capability ? If so, then you might be able
    to convert your 1 gig shadowset into a 10gig shadowset once all the 1
    gig drives have been removed.

    If the disks are not currently shadowed, you will need to dismount/mount
    to create it, and before your temporary licences expire, you'll need
    to dismount the shadowset and remount your disks normally.

    You'll have to stop/restart your applications though because the device
    name will have changed at each dismounT/mount even if the logical name
    hasn't changed (think opened files).

    If you were already on shadowsets, then your plan would work perfectly
    and applications wouldn't see a difference since their drive (DSAxxx)
    would never be dismounted.

  5. Re: volume shadowing question

    JF Mezei wrote:
    > Syltrem wrote:
    >
    >> The VMScluster is comprised of 2 Alphaservers and one IA64 server.
    >> When I mount /cluster the shadow member, will all nodes see it
    >> properly (different architecture should not be a problem on VMS) ?

    >
    >
    > Correct.
    >
    >> there something obvious that I do not see ? Is this a workable plan ?
    >> I remember 5-6 years ago, HBVS would not mount members of different
    >> sizes so that will certainly be a problem if it's still the case.

    >
    >
    > If it works, you need to also ask the question of whether your larger
    > disks on the new system will be usable. Say you have 1 gig drives on the
    > old one and you introduce a 10 gig drive into the shadowset. After
    > you've remove the older 1 gig drives, your shadowset will still only
    > have a 1 gig capacity because that is how the drive was initialised when
    > it was included into the original shadowset.
    >
    > Not sure if I read it or dreamt it, but doesn't VMS now have something
    > like dynamic volume resizing capability ? If so, then you might be able
    > to convert your 1 gig shadowset into a 10gig shadowset once all the 1
    > gig drives have been removed.
    >


    $ set volume/size ...

    However, if the disk was not initialized with /limit qualifier, you'll
    first need to $ set volume/limit on it (which expands the [000000]bitmap.sys
    file so there's room to map the new space.) Otherwise, the maximum
    size available to set volume/size is the size of the bitmap (one cluster per
    bit), which is usually not substantially larger than it already was.

    Unfortunately, while $ set volume/size can be done online, without
    shutting down applications, closing files or anything else, $ set volume/limit
    requires exclusive access to the disk (make sure there's no current access to
    the disk, dismount, mount privately, set the volume size limit, dismount again,
    mount/system.)

    > If the disks are not currently shadowed, you will need to dismount/mount
    > to create it, and before your temporary licences expire, you'll need to
    > dismount the shadowset and remount your disks normally.
    >
    > You'll have to stop/restart your applications though because the device
    > name will have changed at each dismounT/mount even if the logical name
    > hasn't changed (think opened files).


    Also, you'll probably need to enable shadowing with the appropriate
    sysgen parameters and reboot before being able to mount the single-disk
    shadowsets, and remember to reverse this when done.

    >
    > If you were already on shadowsets, then your plan would work perfectly
    > and applications wouldn't see a difference since their drive (DSAxxx)
    > would never be dismounted.


    Works like a charm. For this reason, when configuring a system that
    has a shadowing license, I usually configure *all* the disks as shadow
    sets, even when they aren't shadowed. It lets me expand the volumes
    by adding a bigger disk and then dropping the old, smaller one, and it
    lets me replace a suspect disk by adding a replacement disk to the shadow
    set and then dropping the failing disk, and it lets me do hot backups
    (we're professionals, don't try this at home! - Jamie and Adam) by
    adding a 2nd disk, waiting for the shadow copy to complete, quiesce
    the system, drop the 2nd disk, resume system activity while backing
    the 2nd disk to tape. (Your system/application has to have a good
    "quiescing" feature built into it for this to be reliable, but it
    reduces the offline time for a backup to a few seconds.)



    --
    John Santos
    Evans Griffiths & Hart, Inc.
    781-861-0670 ext 539

  6. Re: volume shadowing question

    On Sep 13, 2:16 pm, "Syltrem" wrote:
    > Hello
    >
    > We are installing an EVA4000 in replacement of an MSA1000.
    >
    > I was wondering if I could use HBVS to switch my data onto the new storage,
    > online.
    >
    > That is: create a new drive on the EVA4000, mount it as a shadow member of
    > the existing one on MSA1000.
    > When the merge is complete, dismount the member that's on the MSA1000.
    >
    > A possible technical problem:
    > The VMScluster is comprised of 2 Alphaservers and one IA64 server. When I
    > mount /cluster the shadow member, will all nodes see it properly (different
    > architecture should not be a problem on VMS) ?
    >
    > And is there something obvious that I do not see ? Is this a workable plan ?
    > I remember 5-6 years ago, HBVS would not mount members of different sizesso
    > that will certainly be a problem if it's still the case.
    >
    > And I hope HP can lend me HBVS licenses for a week or two... I may still
    > have some for Alpha but certainly not for IA64.
    >
    > Just thinking, looking for feedback
    >
    > Thanks
    > --
    > Syltremhttp://pages.infinit.net/syltrem(OpenVMS information and help, en français)


    Syltrem,

    I gave a presentation on precisely this point at the 2007 HP
    Enterprise Symposium in Las Vegas a few months ago.

    The presentation, "Migrating OpenVMS Storage Environments without
    Interruption or Disruption", is available at
    http://www.rlgsc.com/hptechnologyforum/2007/1512.html

    This scenario, and several related ones, are actually excellent
    arguments for using single member shadow sets routinely.

    - Bob Gezelter, http://www.rlgsc.com


  7. Re: volume shadowing question

    Hi Bob,

    Nice presentation! It may be of help to convince some managers over
    here!

    I have one question, though: why are you using the /SYMBOL qualifier
    with the LD commands?

    Regards,

    Bart Zorn

    On Sep 14, 9:45 pm, Bob Gezelter wrote:
    > On Sep 13, 2:16 pm, "Syltrem" wrote:
    >
    >
    >
    > > Hello

    >
    > > We are installing an EVA4000 in replacement of an MSA1000.

    >
    > > I was wondering if I could use HBVS to switch my data onto the new storage,
    > > online.

    >
    > > That is: create a new drive on the EVA4000, mount it as a shadow memberof
    > > the existing one on MSA1000.
    > > When the merge is complete, dismount the member that's on the MSA1000.

    >
    > > A possible technical problem:
    > > The VMScluster is comprised of 2 Alphaservers and one IA64 server. WhenI
    > > mount /cluster the shadow member, will all nodes see it properly (different
    > > architecture should not be a problem on VMS) ?

    >
    > > And is there something obvious that I do not see ? Is this a workable plan ?
    > > I remember 5-6 years ago, HBVS would not mount members of different sizes so
    > > that will certainly be a problem if it's still the case.

    >
    > > And I hope HP can lend me HBVS licenses for a week or two... I may still
    > > have some for Alpha but certainly not for IA64.

    >
    > > Just thinking, looking for feedback

    >
    > > Thanks
    > > --
    > > Syltremhttp://pages.infinit.net/syltrem(OpenVMSinformation and help, enfrançais)

    >
    > Syltrem,
    >
    > I gave a presentation on precisely this point at the 2007 HP
    > Enterprise Symposium in Las Vegas a few months ago.
    >
    > The presentation, "Migrating OpenVMS Storage Environments without
    > Interruption or Disruption", is available athttp://www.rlgsc.com/hptechnologyforum/2007/1512.html
    >
    > This scenario, and several related ones, are actually excellent
    > arguments for using single member shadow sets routinely.
    >
    > - Bob Gezelter,http://www.rlgsc.com




  8. Re: volume shadowing question

    >
    >
    >I gave a presentation on precisely this point at the 2007 HP
    >Enterprise Symposium in Las Vegas a few months ago.
    >
    >The presentation, "Migrating OpenVMS Storage Environments without
    >Interruption or Disruption", is available at
    >http://www.rlgsc.com/hptechnologyforum/2007/1512.html
    >
    >This scenario, and several related ones, are actually excellent
    >arguments for using single member shadow sets routinely.
    >
    >- Bob Gezelter, http://www.rlgsc.com



    Thanks Bob, very interesting !

    Syltrem



+ Reply to Thread