from Hacker News

D2iQ (formerly Mesosphere) sunsetting DC/OS and focusing on Kubernetes

by hartem_ on 11/1/20, 3:24 AM with 34 comments

• by asim on 11/1/20, 12:41 PM

  This is one where the market wants what the market wants and it wasn't mesos. In 2013/2014 I was super excited about mesos and then kubernetes came out. Having worked at Google, seeing this thing come out it was clear from the get go it would win. But if you're a startup with millions in funding running at 100mph you can't stop. It's impossible to course correct immediately and you shouldn't. You have to see it play out in the market. Even in hopes that you can continue to establish a strong business. In this case kubernetes ate everyone's lunch including mesosphere. It's sad. It's really sad to be a technologist who works on something for years only for that to become irrelevant when something comes out that takes the market in a different direction. And this gets even harder when every cloud provider starts using open source as a marketing and distribution channel. I experienced this with grpc. My original project is now no longer a company maintained effort but instead a side project once more and it's even been renamed. The perils of going up against Google. The only saving grace is to solve the same problem but in a different way. E.g move on to being a "platform".

  I'm sad for Ben, Tobi and Flo but that's the game and sometimes it doesn't go your way.

• by nemothekid on 11/1/20, 6:05 AM

  Mesos was always simpler to deploy and wrap my head around, but Kubernetes obviously won the most mindshare. DC/OS was Mesosphere's "enterprise" offering - however I wonder what this means for the Mesos product. I know there are still a couple large organizations still using it, but D2iQ employs most of the engineers working on Mesos, which isn't a good sign if they are no longer making money from it.

• by tnolet on 11/1/20, 8:09 AM

  Sad. DC/OS certainly had the best looking and easy to use UI.

• by dimitar on 11/1/20, 5:46 PM

  "In fact, our recent survey report found that 89% of organizations are running Kubernetes in production or pre-production environments and that 77% of organizations feel that Kubernetes is a central part of their digital transformation strategy."

  Self-selection bias? Of course by "organisations" they mean their enterprise clients.

  I wonder how prevalent Kubernetes truly is, these press releases make it look like it is everywhere.

• by hardwaresofton on 11/1/20, 5:29 PM

  This seems like huge news and seems to be somewhat underappreciated/under discussed on HN... What's the over-under on number of months/years that Nomad will be able to compete?

  I can't tell if consolidation in this space is a good thing (we finally get a single, extensible platform with widespread support, consistent APIs/models from vendors), or a bad thing (see: any other market lacking competition), but outside of Nomad and Swarm are there any other large container orchestrators that people put stock in these days?

  Are we seeing the evolution of the VC model (subsidize-then-dominate) and/or it's application to F/OSS software? If I make a container orchestration platform (I've been sketching one out lately), do I have any hope in ever competing with Kubernetes for attention from developers, no matter how good what I make is (assuming ideological agility could offer benefits greater than millions of dollars is already a stretch).

• by cthalupa on 11/2/20, 2:24 PM

  Mesos was exciting when it came out - the ideas weren't new, but it offered a level of polish and cohesion for those ideas that wasn't readily available beforehand. I'm glad it existed, and glad I got to implement a few projects using it.

  But, I'm not surprised. K8s is easier to work with if you're a startup and really working from a greenfield perspective, and most enterprises still have a lot "Pet" style servers around. It seems like they never found the great fit they would need to survive in the face of stiff competition and a quickly evolving marketplace.

  I wish them the best of luck with this shift and anything else they might pursue in the future. Focusing where they did at that point in time was a gamble, and the type of gamble that pushes innovation in tech forward.

• by whatsmyusername on 11/1/20, 5:48 AM

  Mesos was a joy to work with in the early days of containerization. Easy setup, good documentation, few dependencies. It sucks that it just got outclassed over time by ECS, Docker Swarm, nomad, etc.

  Ending up as another 'kubernetes is a pos to setup and maintain' clone is a gasp of an ending.

• by yuppie_scum on 11/2/20, 4:37 PM

  Used to use Mesos at my last job. It was ok, but the tool ecosystem around it just didn't stay up to date with the needs of the industry. Secrets, Ingress, Cron type problems were difficult to manage and we had to write some in-house stuff to handle it. The zookeeper backend was a pain in the ass of course. Upgrades and recovery from master failures sucked.

  I'm not saying Kube is the easiest thing in the world, but in OSS DevOps what you need is an industry standard with a good tool ecosystem. And Kube has that.

• by eunos on 11/1/20, 4:46 PM

  Quick question, based on this, would it be a bad idea if I decide to use Mesos for Spark based scheduler?

• by pjmlp on 11/1/20, 8:23 AM

  I did not cared about Git, Mercurial was my friend, until I got dragged into Git.

  I got disappointed with what Go might have been, and tried to ignore it as much as I could, although still thinking it is kind of ok for C like programs, now with Kubernetes infecting my world, I slowly feel the need to just suck it up with Go.

  This is why one cannot have nice things.