Rachel's Yard| A New Continuation
Update 1: E3-12xx v4 has a lead time of 4 weeks from Asia via ACME, no source via Superbiiz
So I've been looking at Paperspace.io and Sixa.io for virtual desktop solutions, and it does look very appealing. However, is it possible to run it in a home-lab environment?
Currently, I have a XenServer running on my E5-1650v3 node (256GB) and Apache Guacamole for vdi. I do want to enable some sort of GPU capabilities (like for my Jetson TX2 training, but I will use Sixa.io for now; my workstation has a very shitty support for Linux). Looking at XenServer's HCL, and Intel seems to be a promising solution...
However, Intel is being a stupid asshole. E3-15xx v5 are available only in BGA. (WTF?) and E3-12xx v4 are not available anywhere except very special order. I was able to find v5 but only in Asia (none available in North America). What the fuck?
But, the Skull Canyon NUC does have a Iris Pro processor and unofficially supported by XenServer...
Here's the pricing as of the time of the writing:
On Tuesday half of the Internet broke because S3 was fucked. SlugSurvival happened to have a bug that needed a hotfix. However, since Docker Hub uses S3 as the backend, I could not push my images!
Thus, I have to setup my own in my racks of servers.
First, setup the authentication service. Refers to mkuchin/docker-registry-web. You should have a public key and private key ready.
kubectl create a secret for the public key part.
Then, spin up the registry:
This also assumes that you have a Redis running somewhere in the same namespace.
Third, Profit (of course setup services and ingress as well).
As of the time of this writing,
registry:2 corrupts large layers (see my issue) when using 3rd-party S3. You need to use
It is basically a glorified calendar web app written in VueJS (no backend) that basically helps you (students) plan/search your classes better.
Here's a TL; DR page for you.
Well, sometimes the school's AIS is too slow for my likeky. Also, I have always dreamed of being able to enroll classes with ease. However, a typical enrollment process/checking for classes has always been like this:
Finding your classes should not be that hard.
You are wrong.
For starters, where are you getting the data? It was sort of impossible before, until the school rolled out a better interface on PISA, where it uses bootstrap in 2015, and it is actually human readable. Now we can use all kind of crazy DOM parser to find the class data.
Of course, me being me, always write spaghetti code, then fix later, this is how it looks like right now:
Well, don't worry about it, it gets the job done, at least for now. I will use
next() and what not when I actually have time to improve the code base.
No, it does not enroll users automatically.
Well no shit sherlock. It involves student credentials, and I don't want to fuck with that.
Calm down, it will notify you when your classes are opened. Basically, I have a dispatcher and a bunch of workers to poll data from the website, and insert the changes to the database. It does all sort of magical stuff in the background. Allow me to explain:
So yeah, this is sort of a big project in terms of reliability and automation requirements. I do want to talk to the school and see if they want to use this as part of the AIS.
I will update this post when I have more time and more changes made