Is kubernetes worth it if I intend to self host everything?
This might be the best answer I've ever seen to anything.
If you are already intending to use Kubernetes, then you might as well have a separate cluster to host your business critical applications.
I have been working on getting Kubernetes sorted for my self-hosted applications as well, and it’s been solid. The built in health monitoring is really nice when it redeploys dead containers.
I’m currently working on a plan to re-deploy my cluster with some important infrastructure enhancements loosely based around the work done by BanzaiCloud and their Pipeline project, such as Istio, metrics, logging, tracing and secrets via HashiCorp Vault. Which is overkill, but a fun learning experience.
It may be decent practice, but really it would all run just fine on a single docker host if you're the only user. Kubernetes provides redundancy and scalability, but you pay for it in setup time.
I run the majority of my homelab in kuberbetes. Once you’ve gotten over the initial learning curve, it’s a fantastic tool that id recommend to any homelabber. The only catch is the learning curve
I just finished migrating all my self hosted stuff from a Docker compose file to a single node Kubernetes installed with kubeadm today. Only adds overhead to my system, but I went from having a conceptual understanding of k8s to being able to write Kubernetes yaml files in my sleep, so I’m happy I went through the process.
If you’re just getting started with containers, focus on getting practical experience there first. Maybe I’m too small time but docker compose files are all I have ever needed in freelancing.
I find rancher easier to use than native k8s
+1 to rancher, I have several clusters on a single instance hosted in a data center, works great for that prod/lab split while having a single point of management.
From a purely practical standpoint, you don't need anything but docker-compose. It's more than sufficient for running multiple coordinated containers on a single machine.
However, if you want to learn k8s, there's absolutely no reason to not do so.
I noticed you mention in another comment that you think you'll try rancher. Rancher is fine and good, but if you're wanting to learn for employment opportunities, you're better served by learning straight Kubernetes.
This might be the best answer I've ever seen to anything.
If you are already intending to use Kubernetes, then you might as well have a separate cluster to host your business critical applications.
I have been working on getting Kubernetes sorted for my self-hosted applications as well, and it’s been solid. The built in health monitoring is really nice when it redeploys dead containers.
I’m currently working on a plan to re-deploy my cluster with some important infrastructure enhancements loosely based around the work done by BanzaiCloud and their Pipeline project, such as Istio, metrics, logging, tracing and secrets via HashiCorp Vault. Which is overkill, but a fun learning experience.
You can see my current setup documented here.
This is what I need.
More replies
It may be decent practice, but really it would all run just fine on a single docker host if you're the only user. Kubernetes provides redundancy and scalability, but you pay for it in setup time.
I run the majority of my homelab in kuberbetes. Once you’ve gotten over the initial learning curve, it’s a fantastic tool that id recommend to any homelabber. The only catch is the learning curve
More replies
I just finished migrating all my self hosted stuff from a Docker compose file to a single node Kubernetes installed with kubeadm today. Only adds overhead to my system, but I went from having a conceptual understanding of k8s to being able to write Kubernetes yaml files in my sleep, so I’m happy I went through the process.
If you’re just getting started with containers, focus on getting practical experience there first. Maybe I’m too small time but docker compose files are all I have ever needed in freelancing.
I find rancher easier to use than native k8s
+1 to rancher, I have several clusters on a single instance hosted in a data center, works great for that prod/lab split while having a single point of management.
More replies
From a purely practical standpoint, you don't need anything but docker-compose. It's more than sufficient for running multiple coordinated containers on a single machine.
However, if you want to learn k8s, there's absolutely no reason to not do so.
I noticed you mention in another comment that you think you'll try rancher. Rancher is fine and good, but if you're wanting to learn for employment opportunities, you're better served by learning straight Kubernetes.
More replies