csi-driver-smb VS akri

https://github.com/kubernetes-csi/csi-driver-smb. This is the CSI driver we use.

Actual media storage (movies, pictures, anything that lives on my unRAID box) is mounted to each pod that needs it via the SMB CSI driver. I would love to use NFS instead, but even with 4.x, I was running into stale file mount issues. You can see my findings and why I decided to use SMB instead here

What is the recommended way to use S3 as shared storage for media (videos)? Currently I'm using SAMBA network share RWX volumes using this plugin and I would like to switch to S3 compatible service to increase performance and to avoid my current limit set by my cloud provider on that SAMBA server of 10 active connections for SAMBA.

With SMB I'm using the CSI SMB Driver helm chart to deploy it. When creating the persistent volume I'm able to use some mounting options where I have the following included:

Kubernetes has a native solution for many different storage integrations - CSI. There is a CSI driver for SMB as well. After installing the driver you will be able to map your config files from smb server via regular volume mount and config map.

I mean, in short either find a project that includes support for it (I don't know of one), or look at a CSI that does like the SMB CSI driver.

Using a CSI Plugin you should be able to separate the admin side from the user side. I haven't dove into this but here is a plugin that might help - https://github.com/kubernetes-csi/csi-driver-smb

Look into using flex volumes (though depricated) or its successor SMB CSI driver

akri: AdmissionErrors and if it actually mounts something the process cant access the device for some reason.

If you're talking about hardware attached to nodes, Akri might be of interest: https://github.com/project-akri/akri

Fun hands on use-case/story. Trying to connect sensors & actuators sounds fun! It's clear that they need some kind of infra/platform that they can run on-prem, & this feels like a lock.

There's a new Dynamic Resource api in k8s 1.26 that feels potentially well suited to managing their various kinds of node-attached resources in a pretty broad/generic/cloudy way. Taking the cloud from anonymous compute/storage nodes to an all encompassing management system for whatever we have is an interesting next step. I rather doubt we'll see folks like this spend the 3x effort to pioneer use of these abstractions but over time it should start to emerge & become more regular. https://kubernetes.io/docs/concepts/scheduling-eviction/dyna...

There's also a really neat really nice CNCF Sandbox project Akri which performs semi-similar-ish functions as Dynamic Resources. One of the examples they use as a goto is managing a fleet of OVNIF cameras, like, security cameras. https://github.com/project-akri/akri

On another topic, appreciate this article talking about the mixed human/robot workflows. Good to consider! I worked at a great & interesting data-warehouse company a long time ago that had some very complex long running ETL pipelines & limited compute resources (good number of boxes... pre-cloud era!), and a lot of time-sensitivity. The company used Jira to manage the ETL end to end, with custom workflows where work would gate, process, get checked through various stages, with processors updating the ticket as work advanced or to report % complete. Intensely sharp choice not to build it ourselves, to use a tool for tracking work, even though the code (robots) we built were doing 9/10ths the work. Fun to see a similar topic discussed here.

Good problem setup, interesting view here. Would appreciate a little more technical specifics, but still fun to see.

You need something to act as the runner queue to schedule the nodes, so perhaps you could pull them into a k8s cluster and auto-scale oh the load queue. Using something like this to tag the nodes what special devices might be attached on certain nodes https://github.com/deislabs/akri

Some notes on Plex/Emby/Kodi and transcoding. If you want true transcoding with GPU acceleration, you have to have Nvidia GPU or be a k8s device plugin genius. The whole idea of mounting elastic devices in k8s is fairly new and rather complex. In the mean time transcoding is best done on a beefy device with a proper CPU (eg i7) or specifically Nvidia GPU because there are numerous pre-made plugins. I just run Plex and Emby on an old ATX gaming machine without GPU acceleration and it works totally fine. They were barely usable for just me when running on the RPis, wouldn't recommend it unless you can figure out how to mount the correct devices in the pod using a custom raspberry pi device plugin . . . lol good luck! - Arm labs device manager: https://community.arm.com/developer/research/b/articles/posts/a-smarter-device-manager-for-kubernetes-on-the-edge - Deis labs Akri device manager: https://github.com/deislabs/akri - Nvidia GPU plugin: https://github.com/NVIDIA/k8s-device-plugin