Hell yeah, I think people don’t consider swarm enough, especially for Small Business situations where you do want some redundancy, but not all out k8 setup yet. I’ve seen some weird bugs with NAT and general networking issues with overlay networks randomly stop forwarding published ports. But usually rare and fixed with daemon restart or rebooting host. Swarm secrets alone are worth using it. We ran a single node swarm in dev, just to have all the secrets work the same way they do in prod.
That's a really good and useful point, running a single node in swarm just to have secrets working the same way. I haven't used a single node for a long time and I had even forgotten that secrets only work in swarm clusters. You're absolutely right, I said goodbye to passing secrets as env variables and immediately moved to secrets when switched to swarm
Imho I feel that swarm sort of stopped becoming worth it when k3s became easier to install than docker, and when HA database operators like cloudnative-pg became really good. Just make sure to get S3 for backups outside of the cluster. At this point, the main complexity driver of kubernetes is really just image management and yaml verbosity imho. Github (or gittea for self hosting) has much of the image management part covered with github actions and the github image store, which is a better workflow for actually deploying things to production. But Docker lets a new dev quickly build an image and try it out.
@@BosonCollider That makes sense, my DB high available "solution" in docker swarm is not the most optimal, because if the node where the DB container is running goes down, I have to wait for the DB to be scheduled on another host, I mean, for my homelab that's not a big deal, but that wouldn't be optimal in production. On the other hand, in Kubernetes, like you mentioned I use the percona PXC operator which is a true HA and production ready solution. How do you manage the underlying storage across your k3s nodes? I've been wanting to try k3s, but I started with the traditional kubeadm deployment method so now I don't want to move
@@henrikjensen608 large companies definitely not, but I guess there are still a lot of small businesses (without the resources or need to deploy k8s/k3s) and homelabs out there still running swarm. What do you use?
@@AviDarks Not sure why your comment wasn't showing up, the script just logs in to each host and runs a docker ps command but formats the output, moved the script to a public repo, found here: github.com/linkarzu/scripts-public/blob/faaa87e089c2d646ed1b911813dfdac08f51d806/macos/mac/015-dockerPsAllHosts.sh
@@linkarzu containerd + k8s (a small variant like k3s or minikube could suffice). I run a complete 5 node k8s cluster though because I don't value my own time /s
I get you, I run a 3 CP and 3 worker highly available k8s cluster, installed the kubeadm way. Added Neovim and learning Golang to the mix, and I have no life 🤣 Haven’t tried k3s though, but I see is way simpler to deploy
☕ Support me -> ko-fi.com/linkarzu
☑ My Twitter -> x.com/link_arzu
Hell yeah, I think people don’t consider swarm enough, especially for Small Business situations where you do want some redundancy, but not all out k8 setup yet. I’ve seen some weird bugs with NAT and general networking issues with overlay networks randomly stop forwarding published ports. But usually rare and fixed with daemon restart or rebooting host. Swarm secrets alone are worth using it. We ran a single node swarm in dev, just to have all the secrets work the same way they do in prod.
That's a really good and useful point, running a single node in swarm just to have secrets working the same way. I haven't used a single node for a long time and I had even forgotten that secrets only work in swarm clusters.
You're absolutely right, I said goodbye to passing secrets as env variables and immediately moved to secrets when switched to swarm
Imho I feel that swarm sort of stopped becoming worth it when k3s became easier to install than docker, and when HA database operators like cloudnative-pg became really good. Just make sure to get S3 for backups outside of the cluster.
At this point, the main complexity driver of kubernetes is really just image management and yaml verbosity imho. Github (or gittea for self hosting) has much of the image management part covered with github actions and the github image store, which is a better workflow for actually deploying things to production. But Docker lets a new dev quickly build an image and try it out.
@@BosonCollider That makes sense, my DB high available "solution" in docker swarm is not the most optimal, because if the node where the DB container is running goes down, I have to wait for the DB to be scheduled on another host, I mean, for my homelab that's not a big deal, but that wouldn't be optimal in production. On the other hand, in Kubernetes, like you mentioned I use the percona PXC operator which is a true HA and production ready solution.
How do you manage the underlying storage across your k3s nodes? I've been wanting to try k3s, but I started with the traditional kubeadm deployment method so now I don't want to move
Can you create videos how to install docker swarm? thanks
@@AviDarks I'm considering that, it all depends on how well the docker videos I have posted perform. If people respond to them, of course!
Do we still use Swarm???
@@henrikjensen608 large companies definitely not, but I guess there are still a lot of small businesses (without the resources or need to deploy k8s/k3s) and homelabs out there still running swarm. What do you use?
can you share the file ( the sh fiile ) thanks
@@AviDarks which file?
@@linkarzu 015-dockerPsAllHosts
@@AviDarks Not sure why your comment wasn't showing up, the script just logs in to each host and runs a docker ps command but formats the output, moved the script to a public repo, found here:
github.com/linkarzu/scripts-public/blob/faaa87e089c2d646ed1b911813dfdac08f51d806/macos/mac/015-dockerPsAllHosts.sh
Can you create videos about Linux for beginner , I want to use neovim with Linux:))
Sure!
@@linkarzu thank you very much
@@phamcongtoan1399 no worries, you're welcome!
Neither
What do you use?
@@linkarzu containerd + k8s (a small variant like k3s or minikube could suffice). I run a complete 5 node k8s cluster though because I don't value my own time /s
I get you, I run a 3 CP and 3 worker highly available k8s cluster, installed the kubeadm way.
Added Neovim and learning Golang to the mix, and I have no life 🤣
Haven’t tried k3s though, but I see is way simpler to deploy