BCmoney MobileTV

Although I would tend to agree with 6 out of 7 points in the above opinion piece… shipping should not be optional and in DevOps its not a company nor “business” decision rather a delivery team decision (they’re mistaking Agile for DevOps there)… for instance Scrum insists on “potentially shippable” increments each 1-4 week Sprint/cycle, we chose 2-weeks as our timebox, but then the decision is deferred to the Product Owner as to whether or not the company desires releasing whatever has been completed.

7 DevOps Myths Busted
No offense to its author, Sara Miteva (she seems like an ambitious person with a passion for both Marketing & Technology from her recent bootcamps), but just like you would not take Medical advice from a really good painter, one should look to the trenches of IT for technical perspectives and ideas for solutions in terms of both the definition of DevOps, and how to acheive it:

Companies who entrust their delivery teams to build reliable Continuous Delivery & Deployment automated pipelines following the DevOps model that best suits the team, actually achieve single-piece flow (micro-sized, incremental, “non-existent risk” releases) rather than putting arbitrary wait times resulting larger/unpredictable/riskier batch sizes or ill-informed bureaucracy/red-tape in their way (which DORA and similar research initiatives have shown) are arguably > 200x more productive in terms of platform stability and change efficiency (as measured in Lead Times, arguably the most important metric other than not accumulating let alone paying down of Technical Debt):
https://services.google.com/fh/files/misc/state-of-devops-2019.pdf

Sometimes folks removed from the work can reach indecision on whether its safe to proceed so better to remove the “pain of deciding” away and move that decision into the request process not release process, so if its been requested it goes when ready, if no longer needed it gets removed or flagged off….  summary “if its not important enough to ship immediately upon completion, its probably not going to have any noticeable impact on company performance, and so… should we really be working on such things”

As we know sometimes the answer’s still “yes that is worth working on” if only to avoid “Feature Factory-ism” (to make up a term), and of course to take a proactive TechDebt paydown approach whose benefits are the hardest to measure, but who is in a better position to decide those, the team implementing or someone 2-3 degrees removed from the work (or suffering/toil when things inevitably go wrong if TechDebt is allowed to accumulate)…

Some things I think we would pretty much all agree on…
the current code sucks (but “mostly works”)
every line in the entire codebase once started out as a PR at one point or another…  or else it wouldn’t have made it into our git repo
if we had our way, we would change most of that code to be better (readable/maintainble, design patterns, UX/A11Y, Performance, Security, etc)?!

almost anything is better than what we have now…
we’ve all gotten better at coding/reviewing/fixing over the past 4 years
we have a tendency to as Wilson Philips said “Hold On for one more day” and squirrel away code, far too often suffering in silence, leaving big gaps in when the rest of the team can checkout progress and provide feedback
Continuous Integration is more than just an ideal to strive for, its become industry standard practice for a reason… everyone integrates what they’ve got all the time (daily at least, but more frequently is in fact better)
the opposite of “Continuous Integration” is “Sporadic Divergence”
to change the code and make it better, first you need a PR
if a PR is incomplete, we trust the reviewers to flag any of the more concerning imperfections, so that additional things can be improved… should be considered bonus if PR submitter points out some areas themselves they already know need further improving.

NOTE: this is a re-post of an original article by Kevin Brandon which disappeared but needs to be preserved as it changed many teams’ perceptions of Gitflow usage in relation to Agile

If you’re in technology and haven’t been living under a rock you’re aware of Git. If you’re aware of Git then you’re probably aware of Vincent Driesson’s eventful blog “A Successful Git Branching Model” that has since been implemented and widely referred to as Gitflow. Don’t let the title of this blog fool you, I don’t have a beef with Gitflow. Vincent was accurate to claim he had a successful Git Branching Model. It gets you there 99% of the time for the majority of teams. Still, I have found as far as a team’s ability to be agile, the prescribed flow can run you a foul when indecisive chickens come home to roost.

In Agile the chickens (the customers and executive management) can sometimes interfere (that’s life/reality). For the sake of this post, I am speaking to when executive management promises a feature that was demoed and they want it out yesterday. The problem is they don’t like the other feature(s) that were also in the build. They want it now, if not yesterday. You could argue semantics of implementing agile here; but, at this point “Agile” has left the building and we need to be agile in the original definition of the word.  In my experience it is at this point you suck it up and do as you’re told to those whom provide you a paycheck.  So you set off to get an artifact of exactly what the chickens want.

Well, if you have been exercising Gitflow, you had N+1 number of feature branches and of which several have been reviewed and approved into the develop branch. The problem is after that demo, the big chicken in the corner laid and egg and said “LOVE IT” but I don’t like the N-1 other features. Your product owner is crumbling and they are chanting what that chicken wants. Now you are faced with cherry picking commits and analyzing merges with unwanted features.

A successful solution to this problem involves some slight modifications to Gitflow. I don’t know if I can be as so brazen to say maybe Git Flow 2.0? But this has proven itself to work and if it helps the pigs, all the better!

Several big differences here:

• No notion of a “future feature branch”. Let’s face it, all features are future branches until someone in a higher pay grade blesses it to life. Until then any mitigating/mingling with another feature is a risk of feature entanglement.
• Master branch is back to being the source of origin. Gitflow made develop a first class citizen. develop really is a place we want to have immediate integration of our features. We want to have immediate feedback of our conflicts and a place to demo.
• develop in this flow is best to be automatically merged once the feature branch is mergeable and passes all unit tests.
• Special mitigation branches “develop/proj-1” in orange. Here the feature/Proj-1 had a conflict. Rather than resolving the conflict in the develop branch or feature/proj-1 branch, the mitigation is stored separate for future reference. Once the mitigation is made it is pulled into develop. This is key to keeping your features de-tangled.
• develop should never be directly updated by anyone other than your CI server.
• develop can be at any time destroyed and rebuilt from available features.
• Feature branches are only branched from master, never from develop. This is important again to keep us premature exposure to another feature.
• Releases are made from approved features… not the develop branch… not integration.

Overall this modification to gitflow provides immediate integration feedback, firewalls CI from the world, keeps features detangled, your code review doesn’t inhibit integration, and lets you create releases from chicken approved features. Now that’s keeping Agile to its original roots of being an adjective!

UPDATE (April 20, 2021): folks have continued to consider more streamlined DevOps code delivery workflows, to the point that Vincent Driesson himself (who coined the term and approach for “gitflow”) has all but disavowed its strict adherence in an Agile setting with his May 2020 update, suggesting consideration of trunk-based development, Enhanced gitlfow and/or further simplified “GitHub feature branch” flows.