Automated Multi-Repo IBM App Join Enterprise BAR Builds

The IBM App Join Enterprise (ACE) toolkit has lengthy been used for software improvement and in addition for constructing BAR recordsdata to be deployed to integration nodes, with the IDE’s capabilities making it comparatively easy to work with functions and libraries contained in a number of supply repositories. The toolkit just isn’t simply automated as such, nonetheless, and it could seem that the supply structure have to be reorganized earlier than automation is feasible: the command-line construct instruments lack among the toolkit’s project-handling capabilities and due to this fact current some challenges when working with advanced supply environments. Regardless of the challenges, this text exhibits a substitute for reorganization, with comparatively little work being wanted to permit for automation to proceed.

Fast abstract: The toolkit presents a digital filesystem based mostly on initiatives, and the command-line equal is to repair up the extracted supply throughout a construct; a working instance is proven under.


Many long-standing ACE and IIB prospects separate their storage of supply code (containing software and library initiatives) from their storage of constructed artifacts (BAR recordsdata), with the BAR recordsdata being constructed by builders from supply after which checked into an asset repository. The construct course of includes importing initiatives into the toolkit (from git, for instance), constructing BAR recordsdata, deploying to an area server, testing to make sure that the functions work domestically, after which pushing the BAR recordsdata to an asset repository (similar to Artifactory).

As soon as the BAR recordsdata can be found within the repository, the deployment pipeline begins, with the BAR recordsdata being deployed to a number of environments in flip: Improvement, High quality Assurance (QA), Consumer Acceptance Take a look at (UAT), Pre-Manufacturing, and Manufacturing are commonly-used names, however completely different organizations select completely different names and variety of phases. In virtually all instances, nonetheless, every stage includes extra validation earlier than the deployment proceeds to the following stage. If QA is the primary stage, then the image may look as follows:

This sample is frequent at the least partially as a result of it offers quite a lot of assurance that the BAR file accommodates functions and libraries that work as anticipated: every stage will cowl some elements of testing to make sure that manufacturing deployment will lead to profitable operation. Many corporations have automated a lot of the later phases, resulting in environment friendly development as soon as the BAR recordsdata have been created, however this doesn’t assist the preliminary effort required on the left facet of the image: the preliminary stage includes handbook effort, and this could cut back agility and improve time to deployment for adjustments.

That is very true for adjustments that have an effect on libraries utilized by a number of functions, as many software BAR recordsdata might must be constructed, examined domestically, after which pushed to the asset repository by builders. A brand new launch of ACE can also set off related effort, as may a brand new fixpack in some instances, and because of this many organizations would favor to automate the sooner phases as properly.

One frequent sample to realize that is to have the BAR recordsdata constructed routinely from supply management, and never have the builders construct them: the automated builds create a BAR file and run unit checks, after which push the BAR file to the asset repository when every part has succeeded. Modifying the image above to incorporate this type of automation may seem like this:

The introduction of an automatic BAR construct part permits rather more fast construct and testing of functions when an ACE launch improve is deliberate or an software is modified, and is way nearer to the industry-standard construct pipelines for different languages. This type of pipeline would additionally work with container builds, the place the BAR file is changed by a container picture containing the functions and libraries, with promotion by the varied phases.

This strategy works properly generally, however can develop into sophisticated when functions share frequent libraries unfold throughout a number of repositories: pulling supply down from the repo and constructing it’s simpler when every part is in a single repository (just like the ACE demo pipeline) and will get more durable when parts of the supply code have to be downloaded from a number of repositories. (Word that whereas it’s potential to distribute shared libraries as binary BARs utilizing instruments similar to Maven, this strategy doesn’t work very properly with static libraries, and this text assumes constructing from supply to be able to cowl the widest vary).

This text won’t try and resolve the debates round “monorepo or a number of repos” (see ongoing web arguments), however does acknowledge that it may be sophisticated to get BAR builds getting in an automatic approach for advanced undertaking units. We’re intentionally ignoring the questions round deployment dependencies (“what if integration functions require completely different variations of a shared library?”) as these are liable to come up no matter how automated the BAR construct course of is perhaps. Even when we ignore these questions, nonetheless, there are nonetheless points across the structure of the initiatives on disk, and these are lined within the subsequent part.

Toolkit Digital Mission Folders

Think about an instance set of repositories with an software and libraries in separate places:

with every part in a separate git repo. This may be seen beginning on the App1 repo and whereas the construction is comparatively easy in that there aren’t any “diamond dependencies” or anything tough, it might nonetheless illustrate one of many key points round filesystem structure.

These repositories are linked utilizing git subcomponents, the place the father or mother repo has a subcomponent pointer to the kid repo at a selected commit level. It’s also potential to make use of Eclipse “Mission Set” recordsdata to hyperlink the initiatives, or the Google “repo” instrument, or git subtrees, however the primary rules are related from an ACE perspective, at the least so far as the toolkit is anxious. The Eclipse git plugin handles subcomponents as anticipated, with a “recursive” clone mode much like that of the git command, however ACE command-line instruments instantly begin to run into issues: though the git command can extract the initiatives accurately, when the mqsicreatebar command is requested to construct App1 it fails with errors similar to

Referenced undertaking "SubflowLibLevel1" doesn't exist on the file system

For the toolkit variant of this instance, so long as the top-level software has been cloned with submodules as proven:

then the structure within the toolkit appears utterly regular, with the initiatives showing within the navigator as anticipated:

and BAR recordsdata could be constructed for App1 together with the varied libraries.

Nevertheless, whereas the navigator appears clear and every part works as anticipated, what truly occurs on disk is as follows (undertaking directories in pink):


── ace-submodule-schemalib-level1

│   ── ace-submodule-schemalib-level2

│   │   └── SchemaLibLevel2

│   └── SchemaLibLevel1

── ace-submodule-subflowlib-level1

│   ── ace-submodule-javalib-level1

│   │   └── JavaLibLevel1

│   ── SubflowLibLevel1

│   │   └── SubflowLibLevel1

│   ── SubflowLibLevel1_ContractTest

│   ── SubflowLibLevel1_ScaffoldApp

│   └── SubflowLibLevel1_UnitTest

── App1

── App1_EndToEndTest

└── App1_UnitTest

The toolkit presents a digital structure based mostly on initiatives, no matter the place they might be on disk. Because of this, organizations utilizing the toolkit to construct BARs discover this works very properly, because the toolkit handles all of the complexity. Whereas actuality sometimes breaks by (some schema editor content-assist doesn’t deal with the relative paths accurately), generally this type of undertaking is eminently usable within the toolkit.

Command-line builds, nonetheless, don’t work as properly: each ibmint and mqsicreatebar fail to search out the initiatives within the subdirectories. This isn’t actually a defect within the instruments (somewhat displaying as a substitute how highly effective the toolkit is in presenting a digital undertaking structure!) however is unhelpful in constructing automated BAR construct pipelines.

Reorganizing Initiatives on Disk After Supply Clone

Ideally, the pipeline would look one thing like this:

with the supply checked out at first, adopted by constructing, testing, after which BAR packaging, all with out handbook effort; that is clearly not potential if the command-line instruments are unable to know the filesystem structure.

The answer on this case is to switch the structure on disk in the course of the construct in order that it appears as follows:


├── ace-submodule-schemalib-level1

├── ace-submodule-subflowlib-level1

├── App1

├── App1_EndToEndTest

├── App1_UnitTest

├── JavaLibLevel1

├── SchemaLibLevel1

├── SchemaLibLevel2

└── SubflowLibLevel1

after which ibmint and mqsicreatebar will work as anticipated. Word solely the required the initiatives are moved: the subcomponent checks are left the place they had been, as they aren’t being run at this stage. The subcomponent checks can be run in the course of the construct for these elements, however wouldn’t usually be run throughout software construct (much like library packages in different languages).

To attain this, the instance repo file exhibits how this may be scripted without having to encode the names of the initiatives within the father or mother repository, as forcing the father or mother repository to “know” which initiatives exist within the subcomponent repositories would trigger upkeep points in each locations. Shell instructions similar to

discover ace* -name ".undertaking" -exec dirname ";" | xargs -n1 -i echo mv .

can discover the directories with a .undertaking file in them, and create the transfer (mv) instructions to place them within the appropriate place. Including filters to remove check initiatives is comparatively easy additionally, as could be seen from the script within the repo.

For the shared library construct, an identical strategy additionally works, the place the issue exists however is much less widespread. The listing construction


├── ace-submodule-javalib-level1

│   └── JavaLibLevel1

├── SubflowLibLevel1

│   └── SubflowLibLevel1

├── SubflowLibLevel1_ContractTest

├── SubflowLibLevel1_ScaffoldApp

└── SubflowLibLevel1_UnitTest

 is modified (by one other to


├── ace-submodule-javalib-level1

├── JavaLibLevel1

├── SubflowLibLevel1

│   └── SubflowLibLevel1

├── SubflowLibLevel1_ContractTest

├── SubflowLibLevel1_ScaffoldApp

└── SubflowLibLevel1_UnitTest

with the Java undertaking being moved as much as the proper stage to be discovered in the course of the construct. The Java code itself is the following stage down, however that’s not constructed as an ACE software or library (it’s a plain Java undertaking that’s pulled into SubflowLibLevel1) and so could be constructed and examined with commonplace Java instruments (see the ant build.xml for particulars).

This type of strategy may even work if the repositories are in peer directories (as they’d be for Eclipse undertaking set clones utilizing PSF recordsdata) with minor changes to the scripts. ACE v12 is designed to work with commonplace filesystem instruments so long as the ensuing listing construction is suitable with ACE expectations, so virtually any resolution (even uncommon ones, similar to utilizing docker quantity mounts to create a digital top-level listing) that creates the proper format will work.

Notes on the Initiatives and Jenkins Pipeline

This text is generally about learn how to extract and construct ACE initiatives, and the initiatives (beginning at App1) are designed to point out this side in a easy approach. Every stage additionally has checks to cowl the varied elements of the code, and the checks are described within the repositories for these within the contract check strategy or different testing.

The assorted initiatives are additionally arrange to make use of GitHub Actions to run builds and checks to verify code earlier than it’s allowed to be merged; see the varied .github/workflows directories within the app and library repositories for particulars.

The App1 repo additionally accommodates extra particulars on learn how to set up Jenkins, together with the necessity for superior sub-modules behaviors with the recursive choice set.


Though transferring from handbook to automated BAR builds could seem tough when confronted with advanced supply repositories, the ACE design permits using commodity instruments and commonplace codecs to allow industry-standard pipelines even in advanced instances. No matter whether or not the pipeline vacation spot is integration nodes or containers, the ACE instructions and check capabilities permit for rising agility alongside the entire software program supply pipeline.

Due to Marc Verhiel for inspiration and suggestions for this submit.