I no longer work at Sentry but the complexity of operating Sentry is something that we still worked on when I was there. There is no secret incentive to make Sentry hard to operate. You have to consider that engineers run a version of Sentry on their machines too and any complexity ends up there too.
One of the biggest pain points of running Sentry is the number of containers and a lot of this comes from how Sentry works with Kafka and Rabbit. That pain point is actively being addressed by moving to a virtualized system [1] though I am not sure how long it will take to actually adopt this.
I don't think Sentry is being evil or is trying to make it harder to self host, but I do think that Sentry doesn't have a big incentive to spend a lot of effort making it easier to self host. Again, I don't think that's evil. I think it's entirely reasonable and ok, even if it does make life harder for some people.
Maybe this virtualized system will make things easier. If so, that's great. But if it ends up not working out, or if it does, but over the longer term things get more difficult again, I think that's still just kinda how things happen sometimes, and that's ok.
One of Sentry's goals is for Sentry themselves to operate it as a hosted cloud service. Architecture decisions made to further that goal can naturally and reasonably be at odds with another goal to make it simpler to self host. Sometimes things can't be one-size-fits-all.
I already mentioned this elsewhere here but Sentry does have an incentive to improve this. That’s because the complexity creates challenges in the operation of smaller instances and during development.
> There is no secret incentive to make Sentry hard to operate
It's frustrating when half the comments on a company that dares to open their product is always about how they are obviously intentionally very evil to not do it perfectly/for totally free/with 0 friction/etc.
How entitled have we become lol?
but in this case the order of events was not "awesome product that was open-sourced as-is later" but "OS product that became harder and harder to install over time"
Products get bigger and more complex over time, so "harder and harder to install, configure and manage" is somewhat normal. The money spent on free or "community" edition is very difficult to justify, so it slowly degrades with time until it becomes unusable and people move on to something else. It is part of the normal lifecycle of software.
Hi Armin :-)
I've seen you make this point on similar posts in the past, and I believe that you believe it.
The counterpoint would be that "the purpose of a system is what it does"...
A fairly easy answer is "one of the purposes of Sentry is running Sentry-as-a-service, at large scale". Which comes with compromises that make smaller deployments more "bloated" too, but is a very different thing than intentionally choosing to bloat it to make it hard to run.
It's definitely true that what Sentry is today, is a function of operating it at scale. Thankfully one of the intended purposes of Sentry is also that it scales down much better than today [1]. That it does not is also something that Sentry as a company is not happy with which is why there are desires to change that, which will also benefit self hosted.
[1]: this is because a) developing on sentry has become harder and harder on local machines and b) operating single-tenant installations for some customers brings many of the same challenges that people on self-hosted run into. c) we want more regions of Sentry and smaller ones (such as Europe today) have higher fixed costs than ideal.
Yes.. we're getting there. "one of the purposes of Sentry is running Sentry-as-a-service". Why? To make the dwarves happy? It's making money. And with it comes a whole truckload of very obvious incentives. Notice I don't say anyone is intentionally being evil. That's quite unnecessary.