Endevor

 View Only
  • 1.  Endevor 2.0

    Posted Jun 19, 2019 08:06 AM
    Recently I've been having a think about what are the downfalls to Endevor and have come up with the following statements.

    Statement 1.
    In Endevor I have a 'static' lifecycle, it is not easy to change.

    Why is it not easy to change?:
    - I have multiple repositories (sets of output libraries) for storing my build outputs, one behind each environment in my lifecycle. In-essence this makes an environment physical and not logical.
    - Each environment manages itself with little consideration for other environments (one MCF file per environment). 

    Statement 2.
    In Endevor deployments are limited by my lifecycle. I can deploy to an existing local environment in the lifecycle via Move or to a remote environment via ship. Ship is limited to deploying the contents of a package which in turn is linked to the lifecycle.

    If I was to start again today what would Endevor look like? Personally I would separate out Build, Deploy & Release.

    Build:
    - A flexible & completely logical lifecycle.
    - A single Repository for Storing outputs - separate from the lifecycle.
    - A single database to maintain the relationships between components in the repository, elements and the lifecycle (this is what makes the lifecyce truly logical).
    - The generate process would build outputs, add them to the repository and create the relationships in the database.

    Remember a lifecycle is just a representation of the development process and not a representation of the physical environments. Physical environments can be created, deleted or re-assigned to a different stage of the development process at any time. They can be local or remote and are more frequently becoming less traditional (for example zDT). 

    Deploy:
    - A single mechanism for deployment.
    - Can run multiple processes as part of a deployment (DB2 Bind post process for example).
    - Is a single unit of work (deploy 'this' to 'here' where 'this' uses information from the database to identify the components to deploy and 'here' is the target). 

    Release:
    - The ability to orchestrate deployments.

    Discuss.


  • 2.  RE: Endevor 2.0

    Posted Jun 20, 2019 10:48 AM
    Well, I would agree since what you outlined is pretty much the project scope of the work I am currently doing across multiple LPARs. In fact, your title (Endevor 2.0) is what one of the partitions refers to the work as! 

    A lot of what you discuss is possible with re-architecting and redesigning an implementation today (and WITHOUT a lot of unsupported cutomizations (regardless of source), thank you very much). 

    What has turned into the bane of my existence is twofold:

    1) The desperate need for CA-Broadcom to revisit deployment (vis-a-vis Package Shipment) with an eye towards a complete redesign in order to deliver some true flexibility. 
    2) The seemingly unwavering focus to ignore the growing desperate need to deliver needed functionality in Endevor Classic in favour of a select group of requirements coming from (I assume) a subset of customers that have embraced esoteric things like Eclipse, Webhooks, and RESTful APIs. 

    Don't get me wrong; I firmly believe those things are needed. In fact, Eclipse plays a BIG role in the re-architecting of the site(s) I'm working on. But the reality is the "new generation" of IDEs and the concept of DevOps does NOT plug-and-play with any site that has 20+ years of legacy Endevor customizations! I can't even imagine the chaos that would arise if I introduced Quickedit into that environment, much less Eclipse!!

    ------------------------------
    Consultant
    John D Consulting Inc.
    ------------------------------



  • 3.  RE: Endevor 2.0

    Posted Jun 20, 2019 11:39 AM
    I do keep asking about progress of 'Agile Package Delivery' and have recently added a new ideation 'Agent Style deployments for Endevor'. That said we are one of those sites who do use web services, webhooks and Eclipse IDEs and are eyeballs deep in DevOps. Luckily we're also a sites with few customisations and a fairly robust implementation.

    One of my concerns is that the new REST APIs will hit a wall when Endevor becomes more of a 'back-end' and developers want to build and deploy more freely than our current lifecycle allows. The REST APIs will not be much help if you want to quickly build and deploy to a zDT box for example as the feature to do that style of deployment is not available.