/tech/ - Technology

Ops topics:

Python
Golang
Cloud
Linux
Networking
OpenStack
Docker/Podman
Kubernetes
Devops/SRE in general

crosslinking SRE thread: >>14670

>>21204
>Linux
Only one Linux distro is suitable for server use: Slackware. With every other one you will run into a problem with the package manager sooner or later. And if it is on a production server, that means backing everything up and reinstalling everything, and informing users of the 24h downtime.

>>21207
Just use Debian lol

>>21207
>Only one Linux distro is suitable for server use: Slackware
uyghur are you serious

Most companies use RHEL or a clone (i.e. Alma) or Debian/Ubuntu.

>>21210
>Debian
Good luck with APT pulling in Qt and GTK packages on your headless server.
>>21211
>Most companies use RHEL
Yes, because they can pay someone to manage it for them. For companies is important that responsibility/risk be mitigated/offloaded. For example, if your service is unavailable due to a problem with your OS, and your OS management is outsourced to some RHEL corp, then the fault and therefore liability rests with them, and your customers cannot ask for a refund because the disruption happened through no fault of your own. Why RHEL? Because it is an established foundation that already has all the training and certifications in place so it is easy to train or hire the right people. Ubuntu has done the same, that is why Ubuntu Server is also popular.

I'm sure Slackware servers/hosting/infrastructure exist, but it is probably done in-house and for historical reasons. I'm sure there are fewer of them now, ever since RHEL and Ubuntu became solid enough to actually use.

For a server infrastructure you run yourself Slackware is perfect. There are no "automatic updates". You don't just click through the info box to install the latest updates. If everything is working, then your computer does not change between shutdowns and reboots. The updating is done by you and only of the programs you wish to update. If the only change between versions of a program on a headless server are Wayland/X bindings, then why should you update and risk potential problems when you don't need those Wayland/X bindings?

Where companies relegate OS/package management to a company like RHEL, you, by choosing the same tools, are relegating those responsibilities to potentially hundreds of random people. And you have to trust that when you do apt-get update && apt-get upgrade your system won't break.

Speaking of systemd, look at all the shit systemd does at boot, as PID 1, with a 3m wait for a service to come up before moving on. If you're managing a server remotely (without that fancy server management hardware that allows you to see the status of a machine remotely) it could take 20-30 mins to boot. If it boots at all. With slackware, there is no mystery what is going to happen after reboot.

I'm starting to think there aren't that many adults on this website who know how the world works.

>>21212
>Good luck with APT pulling in Qt and GTK packages on your headless server.
apt-get purge then apt-get autoremove

>>21213
also you can just check package dependencies before installing and use the proper switch (-) to ignore certain ones

>>21212
whats stopping someone from making slackware enterprise edition and making a company based on that?

>>21215
MONEY

>>21219
why didn't that stop RHEL/Ubuntu though? especially because they came later

>>21212
>Good luck with APT pulling in Qt and GTK packages on your headless server.
>2023
>caring about servers
uyghur i just dump docker images on clusters now. no need to manage packages anywhere except in my image which can be easily purpose built to be as minimal as needed.

>>21242
not every company has gone to the cloud anon - plenty of people are still doing VMs with linux and configuring them via ansible. Hell, some companies are still running mainframes, even though its 2023. My corpo has been doing a cloud rewrite for like the last 7 years and they still ain't done.

How do I let the "devops" team at work know that they really are just "ops", without hurting their ego?

>>21250
they probably already are painfully aware anon

AWS Certified Solutions Architect – Associate Certification

pdf study guides for current version as of 2023 (SAA-C03)

opsanon here, Im trying to get a job in devops/sre from from support analyst position. Basically my plan is to pick up some AWS/Kubernetes certs and maybe a red hat cert. Right now im studying for AWS Solutions architect (lower level). I'll let people ITT know how it goes and my progress.

I am a programmer who admins infra from time to time. just old school computers guy who also worked in datacentres and provisioned and maintained cloudshit for companies in addition to writing software.
I hate these non-academic corporate certifications. they largely fell out of favour after the 90s when people realized you don't need someone with a certified sun/ibm/microshit thing to throw together a bunch of servers and switches, anyone who understood networking and consumer grade linux boxes could do it and in fact small smartups built upon consumer hardware and LAMP on top killed off the bloated corps being run by certified solaris(tm) with java(tm) guys.
now we are seeing this stuff rise again because those same companies that killed the old ones (google, which was built upon cheap linux stuff) now have their own "stacks" to hawk (gcp, like aws) and selling you the same corporate seals.

I hope stuff like kubernetes and openshift makes these cloud products as widespread as basic web stacks and we no longer have to suffer this corporate certification grift.

>>21263
>openshift
don't you mean openstack? Plus learning AWS/GCP isn't that bad since almost all clouds including private ones based on for example openstack have similar offerings.

Some companies follow the model you lay out there, ex: Canonical uses openstack/kubernetes

>>21249
>cloud rewrite
generally not a good idea imo, unless your rewrite is cloud-agnostic.
for example, app depends on a message queue and can interface with say amazon SQS for that, that's great. but app that integrates directly with amazon SQS (and S3, api gateway, whatever) is a recipe for vendor lock-in and being sucked dry by the cloud provider.
that means maybe you can't use every little feature of a cloud provider's product but imo we should have learnt this lesson about lock-in after what happened in the 90s. aws is the new ibm, google cloud is the new sun, azure is the new and improved microborg.

>>21286
yeah sorry, openstack.
there's nothing wrong with learning how to maintain cloud infrastructure. my complaint is mainly against the certification grift which over time seems to become a requirement to apply for jobs. they aren't cheap and like you said gcp/aws are similar anyway, so if you know one you will manage to work well enough with the other too, but a lot of jobs will specifically look for a gcp certification because they're a gcp shop, and filter people who know how to use aws, which is silly.

>Canonical uses openstack/kubernetes
yeah, and that stack is open source, so you're dealing with better compatibility between cloud providers who host with that and standardized tools. less chance of lock-in.

>>21288
>is a recipe for vendor lock-in and being sucked in
In practice there's no real alternative to switch to anyway, for certain use cases.

If you want to, for example, create a video streaming service similar to twitch or youtube theres no way you can compete with the variety of edge locations + local zones amazon has and CDN to deliver your video that much faster plus they even have shit like amazon wavelength to embed edge servers for CDN directly in 5G cell towers without ever leaving the cell network. There's no way in hell you can compete with that using private data centers and to even come close you'd have to spend billions around the world. The closest would be maybe use cloudflare or another third party service for CDN. End users don't care about open source they just want their tiktoks and porn streamed 0.1 seconds faster.

https://cloud.google.com/kubernetes-engine/kubernetes-comic

pretty good comic explaining kubernetes

>>21290
that's just for cdn though, with cdn you aren't really integrating as much as you're just pushing content to it. if tomorrow you decide to change cdn provider, you don't have to do a massive rewrite, just change the library that pushes your content.
I mean you don't have to tightly integrate with other stuff like using cloud-provider specific db extensions or embedding their monitoring code into your software or write your api definition in their weird format, etc.

>>21484
cdn was just kind of one example, the rest of their shit is pretty good too, managed DBs eliminating the need for DBAs and capacity planning, and the syntax isnt really different since aurora is just mysql/postgres under the hood. I mean the fact that its just a one stop shop with all these managed services and DNS or whatever. Its kind of just academic unless you are running an extremely edgy site (ex: 8chan) and expect to get deplatformed by AWS

pdf dump from SRE thread

went back to school last month for a network and cloud degree, this first semester is super basic, any recommendations for networking related things I could be doing on the side to learn a bit more and have fun with?

>>21628
how long is the degree? how many years?

>>21629
around 2 years, just an associate's

>>21628
sign up for one or two of the free cloud accounts. i think they just need a different email.
they give you some initial credit. use it to build virtual networks, routing, monitoring stuff, firewalls, ipsec vpns between accounts, gateways, etc. I think that could help.

>>21631
>>21630
>>21629
If your degree is only two years you may as well study for a few certifications as they will increase you employability. Shit like A+, Network+, Linux+, Security+ etc. If you have to choose one focus on Network+.

However I must warn you that its often the case that employers want a four year degree, if theres some sort of program that you can use to convert your associates to a bachelors at a local 4 year college try looking into that (if it exists).

>>21628
>any recommendations for networking related things
You want to read Computer Networking: a Top Down Approach (or Computer Networks by Andrew Tanenbaum). Try to get CCNA cert if you want sysadmin/network admin type of work. For DevOps, you want to learn how to use Docker and basic usage of Ubuntu and Alpine Linux (these 2 distros are often used for Docker). Learning GNU/Linux skills is important if you want to work in anything that's related to computer networking or DevOps. I recommend that you install GNU/Linux (dual-boot Xubuntu or something) if you don't use it already, and read this free book https://linuxcommand.org/tlcl.php

>>21646
>Computer Networking: a Top Down Approach (or Computer Networks by Andrew Tanenbaum)
those are more of a theoretical approach to networking you would have in a CS degree, not a practical guide for network engineers. Its fine to read those but something like Net+/CCNA or other cert.
>For DevOps, you want to learn how to use Docker and basic usage of Ubuntu and Alpine Linux
Devops/SRE certs are mostly about AWS Architect, Kubernetes, and maybe some IAC tool like terraform. Most places are replacing docker with podman. If you learn linux for employability id recommend RHEL or equivalent distro since that covers podman and also most big businesses use it including AWS/oracle/etc. which are just RHEL clones anyway. Ubuntu is mainly for desktops and hobbyists.

Whats /tech/'s opinion on linux certs?

LFCS (linux foundation) vs LPIC-* (linux institute) versus Redhat versus suse/oracle/etc. ??

>>24568
Based and tech-pilled, but what are you getting certified in Linux for? Networking? Get certified in networking first, I've learned more about the Linux kernel from trying to run Lutris than any textbook I've read. As with other certs, it's only worth as much as you can get out of it. If it can promise you the job, go ahead, but I wouldn't vouch on it.

>>24579
>Linux for?
sysadmin cert probably. Just expanding my ops knowledge