Minimal Cluster Requirements - VMS

This is a discussion on Minimal Cluster Requirements - VMS ; All VMS group members, I have a client that is running two separate Alpha 1000A VMS servers, each with its own local set of applications and data. These are old, non-supported Alpha servers that are being used to support local ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: Minimal Cluster Requirements

  1. Minimal Cluster Requirements

    All VMS group members,

    I have a client that is running two separate Alpha 1000A VMS servers, each
    with its own local set of applications and data.
    These are old, non-supported Alpha servers that are being used to support
    local government operations. We believe this is a RISKY thing to do,
    despite the solid, reliable engineering designed into these servers.

    The two Alphas are not clustered. We believe the client should establish a
    cluster that provides a basic level of fall-back and higher availability, in
    case one of the boxes develops a serious problem that potentially could take
    down public works or accounting operations. This will buy them some
    time to figure out what to do next.

    Can someone out there provide an overview of the minimum steps needed to
    establish a cluster with these two boxes?
    Pertinent facts:
    -- boxes are in separate locations about 4 miles apart
    -- fibre exists between the two locations
    -- both boxes are running OpenVMS 7.3 with local OS boot disks
    -- the agency also runs a large Windows-based server farm for Exchange and
    SQL-Server, available on the LAN; some LAN storage is used to
    back up pertinent Alpha files and data
    -- TCP/IP, DECnet and LAT are all running on the network

    I am familiar w/ the 1000A server; I used to run one as a hobbyist at home,
    but I never had it clustered.

    Assuming these two boxes can be fibre-linked together over the WAN and are
    close enough together to keep themselves alive. I also
    assume the disk drives in each box can be "served" to the cluster and "seen"
    by each cluster member.
    But, I'm having a hard time figuring out where to establish the quorum disk,
    since all the Alpha disks are mounted in the server boxes. Do we
    need a third VMS server someplace?

    Could someone please enlighten us as to how to create a basic cluster so
    that these two boxes can be "true" backups for each other?

    If you would, please copy your reply to: lars1227 at verizon dot net
    as well as to the rest of this group.

    Thanks.


  2. Re: Minimal Cluster Requirements

    On Sep 7, 11:16 am, lars1...@verizon.net wrote:
    > All VMS group members,
    >
    > I have a client that is running two separate Alpha 1000A VMS servers, each
    > with its own local set of applications and data.
    > These are old, non-supported Alpha servers that are being used to support
    > local government operations. We believe this is a RISKY thing to do,
    > despite the solid, reliable engineering designed into these servers.
    >
    > The two Alphas are not clustered. We believe the client should establish a
    > cluster that provides a basic level of fall-back and higher availability, in
    > case one of the boxes develops a serious problem that potentially could take
    > down public works or accounting operations. This will buy them some
    > time to figure out what to do next.
    >
    > Can someone out there provide an overview of the minimum steps needed to
    > establish a cluster with these two boxes?
    > Pertinent facts:
    > -- boxes are in separate locations about 4 miles apart
    > -- fibre exists between the two locations
    > -- both boxes are running OpenVMS 7.3 with local OS boot disks
    > -- the agency also runs a large Windows-based server farm for Exchange and
    > SQL-Server, available on the LAN; some LAN storage is used to
    > back up pertinent Alpha files and data
    > -- TCP/IP, DECnet and LAT are all running on the network
    >
    > I am familiar w/ the 1000A server; I used to run one as a hobbyist at home,
    > but I never had it clustered.
    >
    > Assuming these two boxes can be fibre-linked together over the WAN and are
    > close enough together to keep themselves alive. I also
    > assume the disk drives in each box can be "served" to the cluster and "seen"
    > by each cluster member.
    > But, I'm having a hard time figuring out where to establish the quorum disk,
    > since all the Alpha disks are mounted in the server boxes. Do we
    > need a third VMS server someplace?
    >
    > Could someone please enlighten us as to how to create a basic cluster so
    > that these two boxes can be "true" backups for each other?
    >
    > If you would, please copy your reply to: lars1227 at verizon dot net
    > as well as to the rest of this group.
    >
    > Thanks.



    Lars,

    Merging these two systems into a cluster to provide increased
    availability is a good idea, but care is needed to ensure that the
    merge goes smoothly.

    With all of the storage attached to each host, the most serious
    problem will be with data replication. There are a variety of traps
    here, and I have seen clients fall into them (generally the reason
    assistance was asked for).

    Almost any newer generation Alpha hardware will be of higher capacity
    than the existing hardware. A quorum disk will be needed, or one node
    will have to be the "preferred" node (a situation that I do not
    prefer). On the other hand, one (or both) of the existing AlphaServers
    can be used as a Quorum node.

    Some of the issues involved in this merge will be:

    - hardware
    - storage
    - connectivity (copying gigabytes of data between sites is slow; no
    matter how high the bandwidth)
    - incompatibilities between applications
    - file protections and UIC group assignments (if there are conflicts,
    they will have to be resolved by moving the conflicting groups).

    The goal of using a cluster to provide mutual support and
    reinforcement is an excellent one, and it generally can be
    accomplished without impacting either of the user communities
    significantly, if care and caution are exercised in the process.
    retaining someone experienced with such situations and cluster issues
    in general is also a sound idea. Done properly, this can be
    accomplished almost transparently without disruption. It also has the
    potential to be highly disruptive. (Disclosure: My firm, as do several
    others who are active in this forum, does provide consulting services
    of this type).

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


  3. Re: Minimal Cluster Requirements

    In article , lars1227@verizon.net writes:
    > All VMS group members,
    >
    > I have a client that is running two separate Alpha 1000A VMS servers, each
    > with its own local set of applications and data.
    > These are old, non-supported Alpha servers that are being used to support
    > local government operations. We believe this is a RISKY thing to do,
    > despite the solid, reliable engineering designed into these servers.
    >
    > The two Alphas are not clustered. We believe the client should establish a
    > cluster that provides a basic level of fall-back and higher availability, in
    > case one of the boxes develops a serious problem that potentially could take
    > down public works or accounting operations. This will buy them some
    > time to figure out what to do next.


    I would not put these old systems into a two node cluster, you're
    going to find one node can run alone but the other cannot. On these
    systems, at the distances you measured, there's no way to share a
    quorum disk on dual-pathed physical connections.

    If they were co-located, I'd just suggest adding a third node
    (Charon-VAX would do if you need a supported platform, or pick up
    a cheap used system.) But you'll have problems managing the cluster
    common disk when the node its normally on is to be down and when that
    node suffers unexpected outage (VMS may not go down, but how good is
    your power source?).

    If you do decide to cluster, I'd pick up the third node and look real
    hard at making sure that power outage at the site with two nodes
    won't take them both down, as well as looking at redundancy in the
    fibre connection between sites and it's power source.



  4. Re: Minimal Cluster Requirements

    On 09/07/07 10:16, lars1227@verizon.net wrote:
    > All VMS group members,
    >
    > I have a client that is running two separate Alpha 1000A VMS servers, each
    > with its own local set of applications and data.
    > These are old, non-supported Alpha servers that are being used to support
    > local government operations. We believe this is a RISKY thing to do,
    > despite the solid, reliable engineering designed into these servers.


    I'd try to get them to pony up for a couple of used DSxx units.

    > The two Alphas are not clustered. We believe the client should establish a
    > cluster that provides a basic level of fall-back and higher availability, in
    > case one of the boxes develops a serious problem that potentially could take
    > down public works or accounting operations. This will buy them some
    > time to figure out what to do next.
    >
    > Can someone out there provide an overview of the minimum steps needed to
    > establish a cluster with these two boxes?
    > Pertinent facts:
    > -- boxes are in separate locations about 4 miles apart
    > -- fibre exists between the two locations
    > -- both boxes are running OpenVMS 7.3 with local OS boot disks
    > -- the agency also runs a large Windows-based server farm for Exchange and
    > SQL-Server, available on the LAN; some LAN storage is used to
    > back up pertinent Alpha files and data
    > -- TCP/IP, DECnet and LAT are all running on the network
    >
    > I am familiar w/ the 1000A server; I used to run one as a hobbyist at home,
    > but I never had it clustered.
    >
    > Assuming these two boxes can be fibre-linked together over the WAN and are
    > close enough together to keep themselves alive. I also
    > assume the disk drives in each box can be "served" to the cluster and "seen"
    > by each cluster member.


    How are the disks attached? Directly, via HSZs, HSGs, direct SCSI, etc?

    > But, I'm having a hard time figuring out where to establish the quorum disk,
    > since all the Alpha disks are mounted in the server boxes. Do we
    > need a third VMS server someplace?
    >
    > Could someone please enlighten us as to how to create a basic cluster so
    > that these two boxes can be "true" backups for each other?
    >
    > If you would, please copy your reply to: lars1227 at verizon dot net
    > as well as to the rest of this group.
    >
    > Thanks.
    >



    --
    Ron Johnson, Jr.
    Jefferson LA USA

    Give a man a fish, and he eats for a day.
    Hit him with a fish, and he goes away for good!

+ Reply to Thread