I do agree with this article. We are working on a project based on a micro-services architecture and my experience matches.
First of all splinting functionality in multiple services is really hard and the first version of your architecture is probably flawed. It's also really hard to establish the responsibilities of each component, the boundaries, whereas it is really easy to take shortcuts that invalidate the modularity or the re-usability of those components.
Of course, these wouldn't be such big of a deal, except that refactoring becomes really difficult, because refactoring now often involves changes in how these services communicate and moving responsibilities around. Also, we are often talking about teams of more than two or three people, since two or three people will almost always choose to execute a monolith first - so in such teams the responsibilities are often divided between people, with people having an incomplete view of the whole system, so refactoring across the whole stack becomes a real bitch - next to impossible actually if the management or the clients are not acquainted with how software development works, as the development of new features is always preferred over dealing with technical debt (non-software folks do not understand technical debt).
Therefore I agree wholeheartedly with what is being said. It's not that micro-services don't work, however you need very senior people that know how to design such systems and you still have to throw away the first version of the entire system. And if you think you're one of those people that think they get it, but never had a failure, then have some patience, as you'll get there :-)
First of all splinting functionality in multiple services is really hard and the first version of your architecture is probably flawed. It's also really hard to establish the responsibilities of each component, the boundaries, whereas it is really easy to take shortcuts that invalidate the modularity or the re-usability of those components.
Of course, these wouldn't be such big of a deal, except that refactoring becomes really difficult, because refactoring now often involves changes in how these services communicate and moving responsibilities around. Also, we are often talking about teams of more than two or three people, since two or three people will almost always choose to execute a monolith first - so in such teams the responsibilities are often divided between people, with people having an incomplete view of the whole system, so refactoring across the whole stack becomes a real bitch - next to impossible actually if the management or the clients are not acquainted with how software development works, as the development of new features is always preferred over dealing with technical debt (non-software folks do not understand technical debt).
Therefore I agree wholeheartedly with what is being said. It's not that micro-services don't work, however you need very senior people that know how to design such systems and you still have to throw away the first version of the entire system. And if you think you're one of those people that think they get it, but never had a failure, then have some patience, as you'll get there :-)