Prefect in a Docker Swarm cluster - dask

We actually use Docker Swarm to manage our cluster.
Some services are scheduled with swarm-cronjob and run at fixed time, but we want more flexibility.
So we decided to give Prefect a chance and it seems promising, but we can not find how or if it is possible for it to work with warm nodes.
A case would be tu run a task, part of a flow, on a dedicated node of the swarm, identified by a specific label.
Are there natively a support for docker swarm, or some experimentations ?
What would be the best way for you to make prefect and docker swarm work together ?
Do we need to use a Dask cluster on top of docker swarm or is there another way ?

Related

What is a cluster and a node oriented to containers?

Sorry for this question, but I just started with Docker and Docker Compose and I really didn't need any of this until I read that I need to use Docker Swarn or Kuebernetes to have more stability in production. I started reading about Docker Swarn and they mentioned nodes and clusters.
I was really happy not knowing about this as I understood docker-compose:
Is that I could manage my services/containers from a single file
and only have to run several commands to launch, build, delete, etc.
all my services based on the docker-compose configuration.
But now the nodes and cluster have come out and I've really gone a bit crazy, and that's why if you can help me understand this next step in the life of containers. I've been googling and it's not very clear to me.
I hope you can help me and explain it to me in a way that I can understand.
Thank you!
A node is just a physical or virtual machine.
In Kubernetes/Docker Swarm context each node must have the relevant binaries installed (Docker Engine, kubelet etc..)
A cluster is a grouping of one or more nodes.
If you have just been testing on your local machine you have a single node.
If you were to add a second machine and link both machines together using docker swarm/kubernetes then you would have created a 2 node cluster
You can then use docker swarm/kubernetes to run your services/containers on any or all nodes in your cluster. This allows your services to be more resilient and fault tolerant.
By default Docker Compose runs a set of containers on a single system. If you need to run more containers than fit on one system, or you're just afraid of that system crashing, you need more than one system to do it. The cluster is the group of all of the systems (physical computers, virtual machines, cloud instances) that are working together to run the containers. Each of those individual systems is a node.
The other important part of the cluster container setups is that you can generally run multiple replicas of a give container, and you don't care where in the cluster they run. Say you have five nodes, and a Web server container, and you'd like to run three copies of it for redundancy. Instead of having to pick a node, ssh to it, and manually docker run there, you just tell the cluster manager "run me three of these", and it chooses a node and launches the container for you. You can also scale the containers up and down at runtime, or potentially set the cluster to do the scaling on its own based on load.
If your workload is okay running a single copy of containers on a single server, you don't need a cluster setup. (You might have some downtime during updates or if the single server dies.) Swarm has the advantages of being bundled with Docker and being able to use Docker-native tools (docker-compose can deploy to a Swarm cluster). Kubernetes is much more complex, but at this point most public cloud providers will sell you a preconfigured Kubernetes cluster, and it has better stories around security, storage management, and autoscaling. There are also a couple other less-prominent alternatives like Nomad and Mesos out there.

How to set up a cluster in Docker Swarm?

I need to setup a cluster of nodes using Docker Swarm. When I say a node, it should have a db, few apps/services and they need to commuicate with each other. I can create a single instance of this using Docker Compose.
But, I need multiple instances of this setup. So, would Docker Swarm help?
When I say a node, it should have a db, few apps/services and they need to commuicate with each other
Docker Swarm is a tool for managing cluster of nodes, when node is a machine (VM/physical) with docker installed. it's not clear what you expect "node" to be, but with swarm you can run any number of containers which will be deployed across the different machines in the cluster.
But, I need multiple instances of this setup. So, would docker swarm help?
You can define anything on the cluster, but basically swarm is meant to enable communication between the nodes. if you want all your services (which you call "node") to be on the same machine anyway, so swarm will not give you any benefits. just copy-paste the docker compose file between all your nodes. But this isn't something common/recommended in general so maybe if you'll explain better your needs, it would be easier to help you.

Docker Swarm for managing headless containers, and keeping them updated (or watchtower?)

I've been trying to devise a strategy for using Docker Swarm for managing a bunch of headless containers - don't need load balancer, exposing any ports, or auto scaling.
The only thing I want is the ability to update all of the containers (on all nodes), if any of the images are updated. Each container running will need to have a specific --hostname.
Is running docker service even viable for this? Or should I just do a normal docker run targeting specific nodes to specify the --hostname i want? The reason I'm even asking about docker service is because it allows you to do an update (forcing an update for all containers if there are updated images).
Was also thinking that Docker Swarm would make it a bit easier to keep an eye on all the containers (i.e. manage them from a central location).
The other option I was looking at was watchtower, to run on each server that is running one of the containers, as an alternative to swarm. My only issue with this is that it doesn't provide any orchestration, for centralized management.
Anyone have any ideas of what would be a better option given the scenario?
Docker swarm does not give you any advantage regarding rolling updates apart from the docker service command, swarm only provides the user horizontal scaling and places a load balancer in front of those replicas called "service", as well as some other goodies such as replicating the docker events across the swarm nodes.
docker service --force would work as expected.
However, you should probably use both, docker swarm for orchestration and watchtower for rolling updates.

Difference between Minikube, Kubernetes, Docker Compose, Docker Swarm, etc

I am new to cluster container management, and this question is the basis for all the freshers over here.
I read some documentation, but still, my understanding is not too clear, so any leads.. helping to understand?
Somewhere it is mentioned, Minikube is used to run Kubernetes locally. So if we want to maintain cluster management in my four-node Raspberry Pi, then Minikube is not the option?
Does Minikube support only a one-node system?
Docker Compose is set of instructions and a YAML file to configure and start multiple Docker containers. Can we use this to start containers of the different hosts? Then for simple orchestration where I need to call container of the second host, I don't need any cluster management, right?
What is the link between Docker Swarm and Kubernetes? Both are independent cluster management. Is it efficient to use Kubernetes on Raspberry Pi? Any issue, because I was told that Kubernetes in single node takes the complete memory and CPU usage? Is it true?
Is there other cluster management for Raspberry Pi?
I think this 4-5 set will help me better.
Presuming that your goal here is to run a set of containers over a number of different Raspberry Pi based nodes:
Minikube isn't really appropriate. This starts a single virtual machine on a Windows, MacOS or Linux and installs a Kubernetes cluster into it. It's generally used by developers to quickly start-up a cluster on their laptops or desktops for development and testing purposes.
Docker Compose is a system for managing sets of related containers. So for example if you had a web server and database that you wanted to manage together you could put them in a single Docker Compose file.
Docker Swarm is a system for managing sets of containers across multiple hosts. It's essentially an alternative to Kubernetes. It has fewer features than Kubernetes, but it is much simpler to set up.
If you want a really simple multi-node Container cluster, I'd say that Docker swarm is a reasonable choice. If you explicitly want to experiment with Kubernetes, I'd say that kubeadm is a good option here. Kubernetes in general has higher resource requirements than Docker Swarm, so it could be somewhat less suited to it, although I know people have successfully run Kubernetes clusters on Raspberry Pis.
Docker Compose
A utility to to start multiple docker containers on a single host using a single docker-compose up. This makes it easier to start multiple containers at once, rather than having do mutliple docker run commands.
Docker swarm
A native container orchestrator for Docker. Docker swarm allows you to create a cluster of docker containers running on multiple machines. It provides features such as replication, scaling, self-healing i.e. starting a new container when one dies ...
Kubernetes
Also a container orchestrator. Kubernetes and Docker swarm can be considered as alternatives to one another. They both try to handle managing containers starting in a cluster
Minikube
Creating a real kubernetes cluster requires having multiple machines either on premise or on a cloud platform. This is not always convenient if someone is just new to Kubernetes and trying to learn by playing around with Kubernetes. To solve that minikube allows you to start a very basic Kubernetes cluster that consists of a single VM on you machine, which you can use to play around with Kubernetes.
Minikube is not for a production or multi-node cluster. There are many tools that can be used to create a multi-node Kubernetes cluster such as kubeadm
Containers are the future of application deployment. Containers are smallest unit of deployment in docker. There are three components in docker as docker engine to run a single container, docker-compose to run a multi-container application on a single host and docker-swarm to run multi-container application across hosts which also an orchestration tool.
In kubernetes, the smallest unit of deployment is Pod(which is composed of multiple container). Minikube is a single node cluster where you can install it locally and try, test and feel the kubernetes features locally. But, you can't scale this to more than a single machine. Kubernetes is an orchestration tool like Docker Swarm but more prominent than Docker Swarm with respect to features, scaling, resiliency, and security.
You can do the analysis and think about which tool will be fit for your requirements. Each one having their own pros or cons like docker swarm is good and easy to manage small clusters whereas kubernetes is much better for larger once. There is another orchestration tool Mesos which is also popular and used in largest size clusters.
Check this out, Choose your own Adventure but, it's just a general analogy and only to understand because all the three technologies are evolving rapidly.
I get the impression you're mostly looking for confirmation and am happy to help with that if I can.
Yes, minikube is local-only
Yes, minikube is intended to be single-node
Docker-compose isn't really an orchestration system like swarm and Kubernetes are. It helps with running related containers on a single host, but it is not used for multi-host.
Kubernetes and Docker Swarm are both container orchestration systems. These systems are good at managing scaling up, but they have an overhead associated with them so they're better suited to multi-node.
I don't know the range of orchestration options for Raspberry Pi, but there are Kubernetes examples out there such as Build Your Own Cloud with Kubernetes and Some Raspberry Pi.
For Pi, you can use Docker Swarm Mode on one or more Pi's. You can even run ARM emulation for testing on Docker for Windows/Mac before trying to get it all working directly on a Pi. Same goes for Kubernetes, as it's built-in to Docker for Windows/Mac now (no minikube needed).
Alex Ellis has a good blog on Pi and Docker and this post may help too.
I've been playing around with orchestrating Docker containers on a subnet of Raspberry Pis (3Bs).
I found Docker-swarm easiest to set up and work with, and adequate for my purposes. Guide: https://docs.docker.com/engine/swarm/swarm-tutorial/
For Kubernetes there are two main options; k3s and microk8s. Some guides:
k3s
https://bryanbende.com/development/2021/05/07/k3s-raspberry-pi-initial-setup
microk8s
https://ubuntu.com/tutorials/how-to-kubernetes-cluster-on-raspberry-pi#1-overview

Docker Swarm Mode service anti-affinity

We are experimenting with docker v1.12 swarm mode using docker service and trying to find a way to ensure containers do not run on the same node. We have three containers and wanted to run 3 docker-engine hosts. When I initially brought them up with a replica of 2 one of the services ended up running both containers on the same node.
For now I'm getting around this by making them global but I was hoping to find a way to do this. I've seen that you can use labels and then create multiple services for the same container and use constraints but was wondering if there is an easier way.
You can use node labels on nodes and service constraints to influence task scheduling to some extent. But for now swarm mode's scheduler capabilities are limited.
There is an open issue regarding your question without a solution:
https://github.com/docker/docker/issues/26259

Resources