Ask HN: Would you watch Infrastructure/DevOps Screencasts?

[joe-stanton]

I've found that many of my colleagues are eager to learn more about modern infrastructure techniques. In an effort to become more full-stack and better support some of our apps in production. This is especially true of developers arriving via Code Schools rather than from a traditional CS background.

Topic Ideas:

  * Docker/Containerisation (ECS)
  * Service Discovery (Etcd, Serf, Consul)
  * Infrastructure as Code (Terraform, Ansible etc.)
  * Monitoring techniques
  * Capacity planning
  * Security best practices

There doesn't seem to be much out there addressing this "need", except https://sysadmincasts.com/ which seems abandoned these days.

Would this kind of screencast be of interest to anyone? I'm thinking I could charge a subscription fee for it. Small bite size lessons. Any further thoughts or topic ideas?

Thanks!

[WestCoastJustin]

Hey, Justin from https://sysadmincasts.com/ here. Ah, not abandoned just hard to find time. It was taking about 2-3 hours per minute of video if you include all the researching and editing. After joining Docker I just did not have any spare time. Want to get back to it eventually.. one day :) It was incredibly rewarding building something that people find useful and I still get emails every few days from people.. saying "Thanks!". If you go down this road -- at least there is that! It also opened lots of doors. In that a resume was almost irrelevant. So, having said all that, you cannot really go too wrong by helping people, learning new technologies, having doors opened for you, etc. Even if you make no money, like I was (compared to what a company will pay you), you can still do really well.

I created kind of a checklist of things that made the site somewhat successful @ https://news.ycombinator.com/item?id=9837727.

Some personal thoughts if you go down this path. It take a long time to build a following (think 1+ years). Make sure you get RSS feed & email list going from day one (this was a big selling point). I went down the subscription path and was making about $1-2k/month (at $14/month). That was after spending a year putting in 60 hour weeks. Transcripts and diagrams came in really handy in driving traffic to the site. Google was really helpful in building a following. Not saying it cannot be done just that you need to really love what you're doing (in that you'll want to give up before you see a return). Do not skimp on production quality. Having highly edited content with premium audio makes these worth watching (that's the differentiator from other content). Just watch railscasts.com and you will instantly see that quality vs some random youtube videos.

Burnout can also be a real thing here. You paint yourself into somewhat of a corning, in that you are charging a monthly free, for something that requires creative juices. I found there to be real pressure to produce new and exciting content, but what if it is not polished enough or up to your production quality control bar? Might be a good idea to have somewhat of a backlog so that you fall back on in the event you have writers block. I had a few episode, where I just could not write for a couple weeks, or the content was just not up to my production standards yet. You'll also find that as you progress into writing something, you'll think of a much better way to tell the story, then you'll want to re-write on a tight timeline. It was brutal having to produce on a timeline like that.

[joe-stanton]

Wow, thanks for the response Justin. Very useful to get your perspective on this. I have some teaching experience (Code Schools etc.) but this would be my first attempt at screencasting. Very useful to hear some of these issues first hand!

[dozzie]

You know what is more important for a infrastructure/devops engineer, previously simply called "sysadmin"? Understanding several compilers and interpreters (and actually learning several programming languages). Unix C API (fork(), exec(), file descriptors, pipes, sockets, and others). Understanding how services daemonize and how they log. Learning how packaging systems work, how to build a new package, and how to install it. Learning what can be read/detected about the system and what does this information mean. Learning how does the networking work (address configuration, resolver, routing, firewall, packet inspection). Traditional networking helpers, along with several protocols carried out by hand. NSS and PAM, and how the accounts work. And many, many more basic things.

I've never seen anybody understanding the basics, who would have any trouble picking up anything that was a fad in the last ten years from its documentation directly. On the other hand, I've seen Docker or Ansible fanboys that couldn't unify accounts across dozen servers in a sensible way, despite their "modern automation" tools.

And screencasts are the second most useless way of conveying technical material (the top one being podcasts). You can't skim through the material, you can't search it, you can't copy-paste it, you can't print its fragments, it's inherently hard to navigate.

[markatto]

I share your sentiment that there are way too many copy-paste "devops" people in the industry these days; I always thought the idea was to hire people with both sysadmin and coding skills, yet a perplexingly high proportion of people I interact with these days have neither. I'd never choose to hire somebody who can't describe a tcp handshake, explain what epoll_wait() does or whiteboard a hashtable, but that's the state of our industry (and, probably, why those of us who can are in such high demand...)

Nonetheless, it seems silly to complain about a lack of knowledge and then complain about somebody trying to share knowledge. I've definitely wanted to lock coworkers in a room with a copy of _The_Linux_Programming_Interface_ before, but maybe a video series on similar topics would be better-received.

[WestCoastJustin]

> And screencasts are the second most useless way of conveying technical material (the top one being podcasts). You can't skim through the material, you can't search it, you can't copy-paste it, you can't print its fragments, it's inherently hard to navigate.

Transcripts with diagrams and code blocks solves this problem. See https://sysadmincasts.com/episodes/47-zero-downtime-deployme... as a good example.

Disclaimer: I'm the author. Also a Docker and Ansible fanboy.

[joe-stanton]

I've learned a huge amount from screencasts such as destroyallsoftware.com, so I think we can agree to disagree on this point. Additionally, I think the success of Codeschool, Code Academy, Pluralsight etc. show that I'm not really in the minority here.

A number of points you mention are possible topics. Although I can't think of a single situation where "understanding several compilers" would have helped me design/maintain/troubleshoot infrastructure I'm responsible for.

But hey, looks like you're not in my target market, and that's ok!

[afarrell]

> Learning how packaging systems work

Do you have a good source on this? When I sat down to do this in fall of 2014, I found myself frustrated with the lack of good tutorials.

[nicollecastrog]

From the sounds of it, you're not his target audience.

[benzesandbetter]

Definitely interested.

btw, I love https://sysadmincasts.com and would love to see more sites like it.

[kevindeasis]

I would watch it.

So, essentially the egghead.io for Infrastructure/DevOps. If you made it cheaper than egghead.io I would sign up.

[tmaly]

Its not as visual as screen casts, but I would go for a podcast. When I have downtime or I am driving, this is my preferred learning medium.

[arrmn]

Which podcasts are you listening for the purpose of learning?

[alistproducer2]

I would definitely be interested in this.

[ShaneOG]

I'm interested

[gglitch]

I'm in.

[devcheese]

Yes.

[i336_]

It would be really cool to find the sweet-spot between "twitch.tv-like free-for-all that eventually devolves like medium did" and "site run by select group of individuals that eventually develop a clique" - somewhere that lets enough people post so nobody gets too burnt out, but not so many people that it's just like the million and one other videos of this sort of thing on YouTube.

Basically find the balance between supply and demand so that most subscribers have watched all of a given uploader's past videos, while new users won't feel like they have 1000TB of old content they need to go through to catch up.

It wasn't until I read WestCoastJustin's comment (https://news.ycombinator.com/item?id=11670868) that I thought that something like this might be run as a top-down project with a chartered (for want of a better word) set of specific contributors - my initial interpretation was a social approach, where many people could upload videos and become well-known for their experience in specific areas (but still following along with a set of themes specific to your site).

What might be really cool is a system of competitions/incentives/rewards that encourage people to submit high-quality content, along with eg funding small events/hackathons/the like. That sounds like it would be a lot of fun!

Besides DevOps and infrastructure I would also highly recommend you add "enough focus on current programming languages to fully comprehend devops from the dev side of the fence" - focusing on how the engine works without driving the car and seeing the scenery will be boring :D (Translation: go maybe one or two steps beyond TodoMVC, but leave it at that. Then people's appetites have been whetted - which is kind of the idea!)

On a slightly less positive note, this concept reminds me of a service that provides screencasting services for programmers; one of the founders (who may have (had(?)) some mental health issues) who began accusing a video uploader of certain actions in a very confusing way. Not drawing any conclusions myself; the comments (go to your HN settings (username, top-right) and turn showdead on to see them all) are over here: https://news.ycombinator.com/item?id=10486476

I first learned about that service a few months before the linked events happened, and I'd initially filed the site away as something that might potentially be fun to use; I'm not sure how I would proceed to use that service now that this has happened, because I wouldn't want to be caught up in a similarly bewildering sequence of events myself.

Since DevOps is a very interesting subject to me, a site like this would fill a definite hole. I can't promise I'd immediately be able to use the site myself (for current specific reasons that may for all I know have changed by the time the site is up) - but the idea sounds really cool.