VxVM 3.0.2 No Good? - Veritas Volume Manager

This is a discussion on VxVM 3.0.2 No Good? - Veritas Volume Manager ; Our local Sun office says Sun is not certifying Volume Manager 3.0.2, that they are waiting for 3.0.3, which is not out yet. They say that there is a serious problem in the DMP driver in 3.0.2 that causes heavy ...

+ Reply to Thread
Results 1 to 6 of 6

Thread: VxVM 3.0.2 No Good?

  1. VxVM 3.0.2 No Good?

    Our local Sun office says Sun is not certifying Volume Manager 3.0.2,
    that they are waiting for 3.0.3, which is not out yet. They say that
    there is a serious problem in the DMP driver in 3.0.2 that causes heavy
    degradation of I/O throughput, at least on A5x00 Photon arrays. So now
    we are in a Catch-22 in that we need version 3.x in order to upgrade to
    Solaris 7, effectively a showstopper.
    Has anybody heard anything about this?!?

    Steve Murphy
    University of Minnesota


  2. Re: VxVM 3.0.2 No Good?

    Nope have not. We just installed 3.0.2 on a new E10000 on the
    recommendations of Sun.

    Steve Murphy wrote in message
    news:38B6C4A9.69708949@umn.edu...
    > Our local Sun office says Sun is not certifying Volume Manager 3.0.2,
    > that they are waiting for 3.0.3, which is not out yet. They say that
    > there is a serious problem in the DMP driver in 3.0.2 that causes heavy
    > degradation of I/O throughput, at least on A5x00 Photon arrays. So now
    > we are in a Catch-22 in that we need version 3.x in order to upgrade to
    > Solaris 7, effectively a showstopper.
    > Has anybody heard anything about this?!?
    >
    > Steve Murphy
    > University of Minnesota
    >




  3. Re: VxVM 3.0.2 No Good?


    "Brian Canada" wrote:
    >Nope have not. We just installed 3.0.2 on a new E10000 on the
    >recommendations of Sun.
    >
    >Steve Murphy wrote in message
    >news:38B6C4A9.69708949@umn.edu...
    >> Our local Sun office says Sun is not certifying Volume Manager 3.0.2,
    >> that they are waiting for 3.0.3, which is not out yet. They say that
    >> there is a serious problem in the DMP driver in 3.0.2 that causes heavy
    >> degradation of I/O throughput, at least on A5x00 Photon arrays. So now
    >> we are in a Catch-22 in that we need version 3.x in order to upgrade to
    >> Solaris 7, effectively a showstopper.
    >> Has anybody heard anything about this?!?
    >>
    >> Steve Murphy
    >> University of Minnesota
    >>

    >
    >

    I agree. We've had many problems with pre-3.0.2 vxvm that were
    resolved by upgrading from 2.5.x. Also, vxdisk
    works more reliably, and vxdmpadm is a life saver. Before
    3.0.2, we had no way to tell ( reliably ) if DMP had actually
    seen the other path. The second path frequently did not show up.
    Using iostat to compare IOPS between the two paths confirms
    that DMP actually does use both paths. I don't know about the
    load balancing thing, but I know for sure that it both
    round-robins and will survive a faulted controller. It has
    not degraded performance whatsoever.


  4. Re: VxVM 3.0.2 No Good?

    James and Brian, what disk vendor/model are you using with VM 3.0.2?

    James Bouressa wrote:

    > "Brian Canada" wrote:
    > >Nope have not. We just installed 3.0.2 on a new E10000 on the
    > >recommendations of Sun.
    > >
    > >Steve Murphy wrote in message
    > >news:38B6C4A9.69708949@umn.edu...
    > >> Our local Sun office says Sun is not certifying Volume Manager 3.0.2,
    > >> that they are waiting for 3.0.3, which is not out yet. They say that
    > >> there is a serious problem in the DMP driver in 3.0.2 that causes heavy
    > >> degradation of I/O throughput, at least on A5x00 Photon arrays. So now
    > >> we are in a Catch-22 in that we need version 3.x in order to upgrade to
    > >> Solaris 7, effectively a showstopper.
    > >> Has anybody heard anything about this?!?
    > >>
    > >> Steve Murphy
    > >> University of Minnesota
    > >>

    > >
    > >

    > I agree. We've had many problems with pre-3.0.2 vxvm that were
    > resolved by upgrading from 2.5.x. Also, vxdisk
    > works more reliably, and vxdmpadm is a life saver. Before
    > 3.0.2, we had no way to tell ( reliably ) if DMP had actually
    > seen the other path. The second path frequently did not show up.
    > Using iostat to compare IOPS between the two paths confirms
    > that DMP actually does use both paths. I don't know about the
    > load balancing thing, but I know for sure that it both
    > round-robins and will survive a faulted controller. It has
    > not degraded performance whatsoever.



  5. Re: VxVM 3.0.2 No Good?

    I sleep better at night knowing that we are running VXVM 3.0.2 on our two new
    Enterprise 10000's. However, there will not be any applications on the E10Ks for
    another few weeks yet. We have encountered several issues with VM 2.5.x which
    appear to be resolved by upgrading to VM 3.0.2. We use A5100/18GB and A5200/9GB
    disk arrays with our E3500's and the E10Ks are using A5100 disk arrays with 36
    GB drives connected via hubs.

    Regards
    Andrew Stafford
    Andrew.Stafford@canada.com

    Steve Murphy wrote:

    > James and Brian, what disk vendor/model are you using with VM 3.0.2?
    >
    > James Bouressa wrote:
    >
    > > "Brian Canada" wrote:
    > > >Nope have not. We just installed 3.0.2 on a new E10000 on the
    > > >recommendations of Sun.
    > > >
    > > >Steve Murphy wrote in message
    > > >news:38B6C4A9.69708949@umn.edu...
    > > >> Our local Sun office says Sun is not certifying Volume Manager 3.0.2,
    > > >> that they are waiting for 3.0.3, which is not out yet. They say that
    > > >> there is a serious problem in the DMP driver in 3.0.2 that causes heavy
    > > >> degradation of I/O throughput, at least on A5x00 Photon arrays. So now
    > > >> we are in a Catch-22 in that we need version 3.x in order to upgrade to
    > > >> Solaris 7, effectively a showstopper.
    > > >> Has anybody heard anything about this?!?
    > > >>
    > > >> Steve Murphy
    > > >> University of Minnesota
    > > >>
    > > >
    > > >

    > > I agree. We've had many problems with pre-3.0.2 vxvm that were
    > > resolved by upgrading from 2.5.x. Also, vxdisk
    > > works more reliably, and vxdmpadm is a life saver. Before
    > > 3.0.2, we had no way to tell ( reliably ) if DMP had actually
    > > seen the other path. The second path frequently did not show up.
    > > Using iostat to compare IOPS between the two paths confirms
    > > that DMP actually does use both paths. I don't know about the
    > > load balancing thing, but I know for sure that it both
    > > round-robins and will survive a faulted controller. It has
    > > not degraded performance whatsoever.



  6. Re: VxVM 3.0.2 No Good?

    We have tested throughput using both ufs and vxfs (3.3.2 with patch) and the results
    appear to support what our local Sun office is saying. And that is that I/O rates
    with VxVM 3.0.2 are roughly double with DMP turned off. Leaving DMP off is not an
    option. Granted, the testing was done only by doing a mkfile and then by evacuating
    a disk, but the results were the same both times.

    We understand that Sun and Veritas are discussing this issue but that no ownership
    or acknowledgement of the problem has materialized, much less a resolution.

    Andrew Stafford wrote:

    > I sleep better at night knowing that we are running VXVM 3.0.2 on our two new
    > Enterprise 10000's. However, there will not be any applications on the E10Ks for
    > another few weeks yet. We have encountered several issues with VM 2.5.x which
    > appear to be resolved by upgrading to VM 3.0.2. We use A5100/18GB and A5200/9GB
    > disk arrays with our E3500's and the E10Ks are using A5100 disk arrays with 36
    > GB drives connected via hubs.
    >
    > Regards
    > Andrew Stafford
    > Andrew.Stafford@canada.com
    >
    > Steve Murphy wrote:
    >
    > > James and Brian, what disk vendor/model are you using with VM 3.0.2?
    > >
    > > James Bouressa wrote:
    > >
    > > > "Brian Canada" wrote:
    > > > >Nope have not. We just installed 3.0.2 on a new E10000 on the
    > > > >recommendations of Sun.
    > > > >
    > > > >Steve Murphy wrote in message
    > > > >news:38B6C4A9.69708949@umn.edu...
    > > > >> Our local Sun office says Sun is not certifying Volume Manager 3.0.2,
    > > > >> that they are waiting for 3.0.3, which is not out yet. They say that
    > > > >> there is a serious problem in the DMP driver in 3.0.2 that causes heavy
    > > > >> degradation of I/O throughput, at least on A5x00 Photon arrays. So now
    > > > >> we are in a Catch-22 in that we need version 3.x in order to upgrade to
    > > > >> Solaris 7, effectively a showstopper.
    > > > >> Has anybody heard anything about this?!?
    > > > >>
    > > > >> Steve Murphy
    > > > >> University of Minnesota
    > > > >>
    > > > >
    > > > >
    > > > I agree. We've had many problems with pre-3.0.2 vxvm that were
    > > > resolved by upgrading from 2.5.x. Also, vxdisk
    > > > works more reliably, and vxdmpadm is a life saver. Before
    > > > 3.0.2, we had no way to tell ( reliably ) if DMP had actually
    > > > seen the other path. The second path frequently did not show up.
    > > > Using iostat to compare IOPS between the two paths confirms
    > > > that DMP actually does use both paths. I don't know about the
    > > > load balancing thing, but I know for sure that it both
    > > > round-robins and will survive a faulted controller. It has
    > > > not degraded performance whatsoever.



+ Reply to Thread