So it looks like Rancher 2.0 is out, but it looks like Kubernetes is the word of the day with this release. What’s unclear is continued support for using an environment that is made of Docker Swarm nodes like I enjoy in 1.6.x. I hope the direction is not to get rid of support for other “cloud” implementations, I like Rancher and would rather not have to re-architect everything.
Rancher 2.0 fully embraces Kubernetes as the orchestration engine. I don’t think you’ll see Swarm support in Rancher 2.0. Kubernetes is just the overwhelming choice in the community and industry.
2.0 implements much of the Cattle UX on top of/using Kubernetes instead of implementing our own orchestration system. It does not and will not support Mesos or Swarm.
The vast majority of our user and enterprise interest has always been in Cattle and Kubernetes. And if you look at Docker’s focus in the last year, it’s not on Swarm…
Bad decision not to support Docker Swarm.
The complexity of K8s is becoming a deciding factor between the two, even with a tool like Rancher to manage K8s clusters. Complexity galore!
With Swarm Mode and Traefik one can pretty much accomplish anything;
- Reverse proxy, load balancing, SSL termination, etc… via tags
- Swarm supports storage volumes via plugins
- Swarm supports scaling containers workloads on the fly
- Swarm support secret management
- Self healing and scaling the cluster in a reliable fashion (3 masters for cluster recovery, fault-tolerance)
- Even managing multiple mid-size clusters in Swarm Mode can be accomplished using tools like SwarmPit and Portainer. Both open source and loaded with enterprise features for a small fee.
But of course, the big players want to steer the public/industry in an entirely new direction (K8s).
The past 2+ years have very clearly shown we made the right decision. K8s adoption and features and community have grown quickly; Mesos and Swarm have gone nowhere (and Cattle wouldn’t have either). We’ve done very well and are accelerating; Docker was fire-sold and stripped for parts.
Swarm itself has been minimally maintained and developed for years, and Mirantis is clear that they are primarily interested in k8s. Things they’ve talked about doing have gone nowhere and releases since have barely mentioned it.
Actually Docker Swarm Mode is alive an well.
Prominent software like Traefik 2.x would not bother supporting it as a backend if it was dead.
So I would take your response with a grain of salt.
It would be a shame if the community let’s Docker Swarm Mode die.
Swarm Mode deployments are light-years faster that K8s and the learning curve is practically non-existent compared to K8s which requires development efforts like Rancher. Exceptional work thus far with Rancher but Swarm Mode is ideal for small to mid-size cluster, as I mentioned.
Yes the industry wants to steer the public in an entire new direction (K8s) but the reality is one can accomplish pretty much anything with Swarm Mode and Traefik; reverse proxy, load balancing, scaling workloads, etc…
One can even manage multiple Swarm Clusters using tools like SwarmPit and Portainer.
In any case, thanks for the conversation and effort around Rancher.
The “Rancher (metadata)” entry in that screenshot is for 1.x, which is not at all alive or well; so what traefik supports is clearly not a very meaningful indicator.
I have no problems with Swarm personally, but this is already a foregone conclusion; Swarm and Cattle lost. We fully moved on to backing the winning horse, while docker has half-heartedly supported swarm+k8s for years and tried to convince people everything is fine. Mirantis is not going to change anything here.
BTW our “support” for Swarm wasn’t particularly useful in 1.x either; all it did was let you create the nodes and make a cluster. There was never any workload-level Swarm UI in our product.
So if you want that, yes you should definitely use Portainer. They have a nice UI (that was initially a clone of our Cattle one). But guess what orchestrator they support now, and will be focusing on more and more?