Automic Workload Automation

 View Only
  • 1.  Zero Downtime Upgrade — limitations

    Posted Sep 19, 2017 04:33 AM
    Edited by Jason McClellan Jul 26, 2021 08:51 AM
    Recently we have used the Zero Downtime Upgrade* process to upgrade several of our systems from v11 to v12. During the course of testing, preparing, and performing these upgrades, we learned some things about ZDU that were not previously clear to us:
    • Regardingcompatibility mode:
      • Work processeswill not switchtodialog work processmode while the server is in compatibility mode.
      • When the AE is in compatibility mode, once an AE process running the old version stops, it cannot be restarted.
      • When all AE processes running the old version have stopped, this has the effect of finalizing the upgrade, making it impossible to perform therollbackfunction.
      • This means that if anything happens that causes all the old AE processes to stop — e.g., a system outage, power failure, or program fault — this will make it impossible to start these processes again or to perform a rollback.
    • Regarding therollbackfunction:
      • Rollback does not allow going completely back to the old version. It is intended only fortemporaryoperation using old-version AE processes.
      • Once the DB has been upgraded to the new version, it is not possible to downgrade the DB.
    • Regarding thefinalizestep:
      • The finalize step cannot be performed if certain types of tasks started on old-version AE processes are still active (present in the activities list). These includeC_PERIODtasks, but may also include other types. It may not be possible to completely remove these tasks from the relevant tables if no old-version AE processes are running. (PRB00138607 has been opened about this problem.)
      • This means that if all of the old-version AE processes have stopped, but there are still entries associated with old-version AE tasks in some DB tables, it may be impossible to finalize the upgrade without resorting to manual deletions from DB tables.
      • Because of this, in order to ensure a clean ZDU upgrade, it’s probably advisable to ensure that all tasks started on old-version AE processes have completed and have been deactivatedprior to stopping the last old-version AE process.
    The ZDU documentation does not make all of these limitations clear. (The first two will be documented soon.)

    Would someone from Automic please confirm whether this an accurate description of ZDU’s limitations?

    * I noticed this morning that at least for ARA, Zero Downtime Upgrade was recently renamed to Minimal Downtime Upgrade. (Thanks to Carsten_Schmitz_7883 for pointing out that this documentation page pertains to ARA.)

    updated link by Jason McClellan @Jason McClellan


  • 2.  Zero Downtime Upgrade — limitations

    Posted Sep 19, 2017 05:07 AM

    * I noticed this morning that Zero Downtime Upgrade was recently renamed to Minimal Downtime Upgrade.

    Well spotted!

    When I did a ZDU some days ago, I remember thinking "well, this Zero Downtime thing certainly must be for very large values of zero", so this name change is ... apt. :)

    One of course can't help but note that only the header was changed. The text itself still refers in a multitude of places to Zero Downtime, ZDU, and "you can upgrade your AE system without any downtime".

    p.s. both the inability to roll back the DB update, and the temporary nature of any roll back to a previous version was mentioned to me at AutomicWorld or similar occurences. So I can confirm that I heard it both said at some time right from the source.

    p.p.s. I can also confirm, but this only from trying it and thus annecdotal evidence, that you can indeed not start old server processes anymore once the [MZ]DU update has progressed sufficiently.

    p.p.p.s. The link you gave is part of the ARA guide, yet describes the AE. The actual AE guide still speaks of ZDU at present time, as found here.


  • 3.  Zero Downtime Upgrade — limitations

    Posted Sep 28, 2017 03:35 AM
    I received some feedback yesterday from one of the product managers, on the topic of ZDU versus MDU. He explained that the difference in how the upgrade system is described for AWA compared to ARA is intentional. That is, Automic Workload Automation does indeed make it possible to upgrade with truly zero downtime, but  Automic Release Automation does not. I got the impression that Automic intends to bring true ZDU to ARA in the future.

    So far, I have not received any feedback on the other points raised in my post above.


  • 4.  Zero Downtime Upgrade — limitations

    Posted Sep 28, 2017 10:25 AM
    Edited by Jason McClellan Jul 26, 2021 08:52 AM
    PRB00138607 has been opened for the problem that can prevent the finalize step from completing if certain tasks started on old-version AE processes are still active. When this problem occurs, the following error appears:
    ZERO_DOWNTIME_UPGRADE - 'FINALIZE' not possible. Workprocesses to shut down are not idle yet./Please try again later.
    (This is the problem we suspect may be caused by C_PERIOD tasks.)

    Link updated by Jason McClellan  @Jason McClellan
    ​​


  • 5.  Zero Downtime Upgrade — limitations

    Posted Sep 29, 2017 12:56 PM
    The first two limitations are now documented.


  • 6.  RE: Zero Downtime Upgrade — limitations

    Posted Jul 23, 2021 08:39 AM
    @Jason McClellan: the post above contains a broken CA communities link. Would you please fix it?​


  • 7.  RE: Zero Downtime Upgrade — limitations

    Community Manager
    Posted Jul 23, 2021 12:10 PM
    @Michael A. Lowry  - working on finding the new home for the old salesforce link with the CMGR team.   Thx  ~jm​

    ------------------------------
    Thank you
    Jason
    Broadcom Community Platform Admin, IT
    ------------------------------



  • 8.  RE: Zero Downtime Upgrade — limitations

    Community Manager
    Posted Jul 26, 2021 08:54 AM
    @Michael A. Lowry

    updated link by Jason McClellan @Jason McClellan

    old link
    https://automic.force.com/support/apex/CommunityCaseHandler?id=500b000001H2II8AAN

    New Link for PRB00138607
    https://knowledge.broadcom.com/external/article?articleId=88359



    ------------------------------
    Thank you
    Jason
    Broadcom Community Platform Admin, IT
    ------------------------------