MQ

 View Only

  • 1.  RDQM OS upgrade

    Posted 28 days ago
    Edited by John Hawkins 28 days ago

    Hi folks,

    (asking a lot today - sorry !)

    a customer of mine has been told, by IBM, that they can't have a rolling upgrade of their OS  in their RDQM environment because pacemaker will fail/not cope (sorry, not much more info than that!).

    I can see this:

    IBM MQ Replicated Data Queue Manager Kernel Modules

    Ibm remove preview
    IBM MQ Replicated Data Queue Manager Kernel Modules
    The Replicated Data Queue Manager (RDQM) feature of IBM MQ Advanced relies on a kernel module to provide data replication.
    View this on Ibm >

    which confuses me a little -

    does this say that if they are moving from e.g. RHEL 7->8 then they need to update their pacemaker cluster because the kernels are different ? which suggests that pacemaker is a problem for any solution using it when moving migrating versions of OS - not just MQ?

    thanks for any help 



    ------------------------------
    John Hawkins
    Integration Consultant
    ------------------------------



  • 2.  RE: RDQM OS upgrade

    Posted 27 days ago

    Hi John,

    If I recall correctly there is a procedure to follow when upgrading in the RDQM. It involves suspending the server from the RDQM and then adding it back again. As you are changing the OS (RHEL version), you would have to make sure that the corresponding pacemaker and corrosynch modules are also updated and match the OS compatibility tables for RDQM on the upgraded server.

    See section in documentation on how to upgrade the pacemaker version in the RDQM cluster.

    You can then see if you can correctly resume the upgraded server into the RDQM cluster. Make sure that everything is synchronized on the upgraded server before you proceed to the next server for the upgrade.

    Hope it helps



    ------------------------------
    Francois Brandelik
    ------------------------------

    Original Message


  • 3.  RE: RDQM OS upgrade

    Posted 22 days ago

    Hi, John,

    The link you provide relates to the DRBD kernel modules supplied with RDQM, and these are often a cause of complications when applying OS maintenance.

    However, when upgrading from RHEL7 -> 8 the significant inhibitor to a rolling upgrade of RDQM between these versions is that they use different versions of Pacemaker that are not protocol compatible. Preventing the maintenance and shareing of state across the cluster when there is a mismatch of versions within the cluster.  Additionally, there are some aspects of the stored configuration that are not compatible.

    Therefore, to reduce downtime and provide best options for rollback procedures the guidance we provide in the MQ docs is to create a new cluster on a fresh set of nodes, using backup/restore to bring across messaging data and configuration.  Using this technique it possible to swing across queue managers on a one-by-one basis: https://www.ibm.com/docs/en/ibm-mq/9.3?topic=mm-migrating-rdqm-configuration-from-rhel-7-rhel-8

    I am aware of customers who have done in-situ migration, but needs to be undertaken with the understanding that the cluster needs to be fully dismantled and recreated, with the associated downtime to all queue manager in the cluster and greater complexity of any rollback procedures.

    Regards,

    Adrian



    ------------------------------
    Adrian Dick
    ------------------------------

    Original Message


  • 4.  RE: RDQM OS upgrade

    Posted 22 days ago

    Hi, John,

    While I'm aware they often cause complexities for customers applying OS maintenance, in this instance the kernel modules (as per your link) are not the issue.

    The issue for RDQM when upgrading from RHEL7 -> 8 is that it uses different versions of Pacemaker for the two OS versions. Unfortunately, the two versions of Pacemaker are not protocol compatible, meaning mixed versions are unable to communicate state between nodes within the cluster.  This means it is not possible to achieve a rolling migration as we would typically recommend for, say, applying an MQ fixpack.  Additionally, there are some aspects of the Pacemaker configuration that require rebuilding due to incompatibilities across the versions.

    This page within the MQ Docs provides the recommended sequence for migrating RDQM from RHEL 7 -> 8.

    https://www.ibm.com/docs/en/ibm-mq/9.3?topic=mm-migrating-rdqm-configuration-from-rhel-7-rhel-8

    In order to minimise service downtime, and allow easier rollback, this sequence advises recreating the RDQM group and queue managers on fresh nodes, you can use backup and restore procedures to carry across messaging data and configuration.  I am aware of customers who've done in-situ backup/delete/recreate/restore style migration, but comes at the cost of a full outage throughout the whole procedure and greater complexity to rollback if any issues were to be encountered.

    Regards,

    Adrian



    ------------------------------
    Adrian Dick
    ------------------------------

    Original Message


  • 5.  RE: RDQM OS upgrade

    Posted 12 days ago

    Yes pacemaker is an issue. Kernel mapping is needed also when OS patch are applied; It is a cordinated effort.

    The better way on how we do it is; and takes the least amount of effort.

    1. get new host desired version
    2. install pacemaker and mq
    3. backup old qmgr 
    4. create queue manager with same name on the new box with a different listener IP
    5. ensure HA and move of node happen as expected.
    6. Now restore the backup; and flip the DNS.

    All should work as expected.

    the above can be pre-planned and cut-over time is reduced.



    ------------------------------
    om prakash
    Architect
    NorthwesternMutual
    Milwaukee
    4146611399
    ------------------------------

    Original Message