How to learn servers as fast as possible?

I asked in yesterdays /sqt/ but nobody answered

How can I into servers?
I was just low-level it-tech last week and now I'm in charge of infrastructure of a full company and expected to fill a server room in two months. But I don't know shit about servers.
Can someone recommend any good read?

Attached: YBfGxzg.jpg (480x454, 72K)

Other urls found in this thread:

unraid.net/
capturingreality.com/Product
docs.nvidia.com/grid/6.0/grid-vgpu-release-notes-red-hat-el-kvm/index.html
vfio.blogspot.com/2015/05/vfio-gpu-how-to-series-part-1-hardware.html
twitter.com/SFWRedditVideos

The better question is how do i fuck your mom in the ass with my dick? Quite easily I'd assume

Attached: 1312514080736.jpg (996x3019, 579K)

noob here. is it bad practice to run your site on port 80?

Why would you install Ubuntu and then strip it when you can just install a distro that doesn't come with a DE?

Literally no one will be able to reach your website if you don't serve it on port 80.

Port 443

Port 80 as well on yourwebserver to redirect to 443.

Ignore everyone else, HP offer free training for hardware basics, and they also offer hardware monitoring and break fix services, so when a disk fails you will get a hp tech show up to replace it within a few hours.

For the most basic 'do it all myself' data center admin job you will need GOOD experience in:
Storage
Networking
Hardware
Windows Server (incl. DNS & DHCP)
Backups

And that's really just at a minimum, see if you can find a $10k contract with a decent managed services company for the first year or so to cover your own ass.

Heh. You'll need racks, server blades, and enough cabling to do the job. You'll want to learn about cable routing and cooling, more than anything.

Picking the servers should be as simple as picking a vendor and managing your budget.
IBM, HP, Lenovo. Places to start.

Yeah, go with a contractor, for sure.

You can get cheap servers at labgopher.com, haven't tried myself tho

Attached: IMG_20180905_220439.jpg (867x668, 73K)

work on them thats the only way to learn

convert to islam

everything else will come

Ok guyse I understand I was being unclear.
Let me detail:
I know the hardware side of servers: I know CPUs I know PSUs I know RAM and all that stuff.
Also I know Windows server as I managed some in the past years.
Also I know apache servers on debian and some stuff like that.

Now that I know both hardware stuff AND "OS" stuff, where can I learn how to put them together.

Example:
I have a 4U server with 4 eight cores CPUs, 4 graphics cards slots and 1TB of RAM.
For my company I need at least one windows server, two or three Perforce & Jenkins servers, and also I will need the graphics cards of the server to perform high level calculus.
This makes a lot of questions for me:
Can I run all these servers on the same machine ? And how ? I looked up type 1 hypervisors but I'm not there yet.
Can I use the graphics card without impacting the rest of the servers ?
Should I need one separate blade for each purpose ?

you are totally fucked my dude

Thank you my good sir, I have foreseen this.

HP is expensive af, not to mention their shady vendor lock when hw RAID wouldn't work without approved™ HP® branded seagate HDDs.

> Can I run all these servers on the same machine ? And how ?
You need something with VT-d/IOMMU support. Not necessarily just a KVM-like hypervisor, but it's one of the options. Hyper-V, ESXi if they support it or KVM+LXC/Docker (maybe Proxmox).

First week:
Find out your budget and use cases for everything
Weeks 2 to 4:
Buy a fuckload of cables, more than you think you'll need.
Buy some racks
Buy computer hardware.
Put it all in place
Second month:
Get software up and running and spend time tweaking it.
Get scripts set up etc
Test things over and over and over and over. Getting this wrong could be a CK.

Easy, stop overcomplicating things. 2 months is a long time if you manage your time correctly.

Is there a free Type 1 hypervisor out there ?
Or can you recommend the best and cheapest one?

Thanks but I already understand the concept of testing. My question was more like "how do I put that together" speaking about the OSs, the hypervisor if any, can only one machine do all that I need, how do I use the GPUs without impacting the rest, etc

>Is there a free Type 1 hypervisor out there ?

KVM

> GPU [...] perform high level calculus.

What program exactly will be doing this and on what operating system does it need to run?

>Can I run all these servers on the same machine ? And how ? I looked up type 1 hypervisors but I'm not there yet.
short answer yes
long answer:
you need to first define your actual needs clearly before you know what to run and how
as far as hypervisors go, VMWare is very popular among enterprises, for good reasons. Then you have Microsoft Hyper-V which can be good for you if you have a mostly windows infrastructure, and you are already familiar with Windows server management. But at this stage I would recommend you simply hire a consultant to do the job, that's what they are for. They will get you started and then you can learn about your systems on the job.

we don't nee shitheads like you here kid

>Is there a free Type 1 hypervisor out there ?
yes, Qemu and Xen are free, but difficult to setup, especially in enterprise environments
you could also try unraid, which is probabky more affordable than the big ones for SMEs
unraid.net/

>inb4 arch
Ubuntu server exists too, much better option

port 80 is http. so people can easily reach your website, so its good

port 443 is https, now, you can do a redirect or just enable both, the reason you should force https sometimes, is so you can ensure web browsers trust your website, as nowadays if the website is http only, it'll complain

sure, then have it serve a 301 to your https site

good server knowledge requires experience, so starting with a private server worked for me to learn, then I moved onto a public server.

First, getting apache basics out of the way, www is a subdomain, not required but important to include. DNS like google.com actually just points to its IP address, if you ping google and access its website by its raw IP, thanks to server side work, it redirects to google.com. Try for fun, if you have a domain, create wwwwwwwww.website.com and you will see it will work if you point it to your websites IP, real proof that www is just a subdomain.

Freenom offers terrible, but worthy free domains.

Enable ssh and decide weather you want keys or password or both. keys prevent others from logging in unless they have your key, but if you lose the key you will need physical access to the server or contact hoster support.

Okay now learn the rest, I'm tired typing. Have a nice day

It's for RealityCapture, a photogrammetry software that generates 3D models from pictures. This needs a lot of computing power (at least 1 hour for a simple rock using a GTX 1080 Ti)

Cool, so now we're getting down to specific requirements.

capturingreality.com/Product
> RealityCapture runs on x64bit machines with at least 8GB of RAM,
64bit Windows 7 / 8 / 8.1 / 10 (or Windows Server 2008+), graphics card with NVIDIA CUDA 2.0+ GPU and at least 1GB RAM. The CPU is required to support SSE4.2 or higher.*

It should be possible to do this with a linux/KVM hypervisor (I recommend CentOS aka pleb-RHEL), and search for nvidia vgpu for documentation.

I have not set this up myself but I've had to support tangentially related issues on such configurations.

Also, you could hire a consultant to set this (or some other hypervisor) up as another user suggested.

docs.nvidia.com/grid/6.0/grid-vgpu-release-notes-red-hat-el-kvm/index.html

Look for "Windows Guest OS Support"

Disagree. Use coreos

>*tips bedorah* cancer

The fuck, just install a server distro

Clarify what your employer expects in further detail please.

We currently have several desktop PCs used as server laying here and there. We are moving out in two to three months. In the next building we will have a proper server room.

Employer expects me to make use of this server room replacing all the desktops/servers by actual server blades.
On top of that we will need these blades to participate in the distributed computing of different data (code, graphical objects, etc)
I hope it's more clear.

>On top of that we will need these blades to participate in the distributed computing of different data (code, graphical objects, etc)

First of all:
> (code, graphical objects, etc)

nobody cares and neither should you.
You should care about what the software and hardware requirements of said software are.

> we will need these blades to participate in the distributed computing

Is this a requirement? Does it have to be "distributed"?

Just make a big list of things that are needed.

Also, if you're now in charge (do you feel in charge?) it is expected that you will be shit technically, not because you're a noob, but because all managers are shit at technical stuff.

Your job is to make lists of things that have to be done and then put price tags next to the list items. The price tag is either going to be how much you need to pay someone else to do $thing or how much workhours it will take you to accomplice $thing.

>You should care about what the software and hardware requirements of said software are.
Well the more computing power the faster things will be. We have a deadline approaching and we must always go faster.

>Is this a requirement?
As I said, yes. We must compute faster.
>Does it have to be "distributed"?
"Distributed", "shared", "blockchained", call it however tf you want: everyone of our employee has a "compile" software on his computer and he can compile whenever he wants; when he does so, the compiler of his computer talks with the compiler of all other computers which are then using some of the CPU to help the first one to go faster. This is for the "code" part.
For the "graphical object" part, the same goes but it's with GPUs. (and it's just one guy in charge of this)

> if you're now in charge
I am.
>do you feel in charge?
Yes. I've never done so much hours. Basically I come at work before 8:30, it's 19:30 and I'm still there.

>it is expected that you will be shit technically
Yes they say they understand because I'm a trainee. I'm at school one week per month. But I still need to do today the job I'll be supposed to do in the end of my learning, in two years.

>Your job is to
My job is to do everything. Think about everything. Put everything in place. I litteraly have no one to reffer. People here think Ccleaner is a fucking antivirus, do you understand how fucked I am ? Man I came to Jow Forums for advices, you can guess how desperate I am right now.

Sorry if I sound harsh, I'm exhausted.

Would your bosses not spring for Tesla cards? It is what they're designed for

>apache

using crapache in 2018

> first I start by drawing this head...

And with PCI passthrough, you shouldn't be using consumer GPUs. You can, but you have to do driver hacks (for Nvidia, anyway), so you have to worry about a vendor finding out and deciding they can drop your support contract for "unsupported configuration" BS

>jenkins
i hope you guys already have a set of plugins
if you're deploying jenkins from the ground up...prepare to bang your head against the wall in frustration

qemu/kvm + gpu vfio passthrough is pretty straightforward when it works, with server grade hardware (xeon(s) and quadros) it should be super ezz

link: vfio.blogspot.com/2015/05/vfio-gpu-how-to-series-part-1-hardware.html

We had a general going on about servers /hsg/ should've asked there in time.