What service would be best for auto updating/deploying my daemon across all docker swarm workers
title
23 Replies
Thanks for asking your question!
Make sure to provide as much helpful information as possible such as logs/what you tried and what your exact issue is
Make sure to mark solved when issue is solved!!!
/close
!close
!solved
!answered
Requested by dawson.gg#0
hi @AeonRemnant you know this shit
Don’t use Docker Swarm is the answer here, it’s buggy as fuck and unstable.
I wish it was a valid solution. :/
K8S is the solution :/
Basically, yeah.
what.
you say this now.
LOL ok
shouldn't be too hard
what's a nice way of deploying my daemon's across all nodes with roll backs and all that?
I use coolify atm, I just hate the networking section of it
I've always said it. Swarm is fine only if it's not in prod and you're ok with bad instability.
If you're trying to do things that're even remotely important you use k8s or k3s.
okay well it's in prod 😭
do you have a recommendation for either? i tried to use kubernetes for carbon in the past, didn't go great
IIRC, k3s was much simpler to setup and made byyyy rancher?
I would recommend you just don't use distributed computing if you're going to try and use 'the easy way'.
Distributed compute doesn't have a good easy way, and it actively tries to footgun you at all junctions if you don't study.
Learn how to k8s, once you can keep a cluster functional and stable for a year in the lab, you're good to begin thinking about prod.
a year? Yea that can't happen
I could pretty easily move back to docker and we handle the orchestration
Well, my other solution for you is to pay someone that knows what they're doing to know what they're doing.
Or setup some custom Go code with normal Docker and a bunch of RESI APIs.
Extremely expensive, and I would want someone long term that shared a vision with me
This is what I had in the past, I might just go back to it
It'll do for the time being.
v2 recode incoming
Just for fucks sake don't touch Kubernetes unless you study first.
A year in the lab, actively use the cluster, if it's stable after that and you've learned how to basic DevOps? Good to small prod.
Yea a year of testing is never gonna happen ill tell you that rn 💀
That's not what I said, I said you need to learn how it works.
"A year in the lab"
that's what I was referencing
You learn in the lab.
That year is so your fuckups don't hurt your customers.
Didn't realize that was something different?
Oh it is.
That year is specifically so you can explore how and why Kube works the way it does and to think about how to architect a solution correctly.
One common mistake new Kube admins make is they rush into a functional architecture too quickly without considering how to migrate away from it if they need to, how to expand upon it if they need to.
Then what happens is project scope creeps up and whoops, the entire project just imploded.
good to know, thank you
That's why I say a year, because after about 4 months of stability you'll go "Hmm, but what if I go from 3 nodes to 6 for extra redundancy?" and because you didn't consider how to scale up your app lifecycles you'll find your database crashes and now you're dealing with a corrupt DB while the cluster drops from under your feet.
Kube is a footgun and nobody sees that until it's too late.
Or the ever classic example of admins not setting limitRanges or resourceQuotas and the instant load hits the server the entire cluster OOMs and goes into a death spiral because they WAY underestimate how fast Kube resources can balloon.