Hey all, I've been a computer hardware enthusiast for a long time but I've never started to explore the enterprise realm until the past twelve months. I've put together a homelab and I want to start playing with VMs for various reasons(PfSense, Docker, ect.). However I'm really having a hard time figuring out where to get started and I'd like to be pointed in the right direction.
I was getting started with just some Hyper-V stuff but I don't particularly like how Windows Server handles hardware, and watching all these videos about IMMU Grouping, pass through and such makes me want to try Linux. I was trying to make a ZFS Root partition with Unbuntu but aside from being defeated by the command line I've also been defeated by not realizing that having a bunch of single drives as individual "pools" provides literally no redundancy like I thought it did(Maybe I'm a little dumb but as a complete newbie I found This video by 45 Drives to be kinda misleading to that degree).
In short, this world is filled with many shiny things and I have no idea where to get started. I'm using a Dell PE R710 for this project, I've got 24GB of RAM and 4x 500GB HDDs. I started with six of them a year ago and two have failed, leaving me with a great desire to have some parity until I can afford new drives.
PS. First post here, I apologize if i broke any rules or placed this in the wrong category!
No he means the actual storage pools, not how the pools are organized. OP that's why fibre channel, iscsi, etc exists. You can create storage pools in there with performance queues, redundancy, hot spares etc then point your hypervisor towards that's.
What do you mean you didn't like the way hyper v handled hardware? Just not having enough low end control?
I guess I am confused cause if zfs is setup with a mirror or raidz then the VMS storage is on top of that and there is your redundancy..Am I missing something?
I had intended to skip over my Raid Controller entirely and just have 4 drives, JBOD, as 4 "pools". However the recent interview video has led me to believe that's not going to have any parity at all. If I wanted I could take those four drives, put them in 2x Raid 1's with two drives each. Then have parity at the RAID level, but having my storage halved then loosing more space to ZFS isn't appealing.
Yeah, it seemed like with some of the legitimate "pass through" stuff in Linux I could just hand over a Ethernet port to a VM, not have to deal with virtual hardware on Hyper-V. If it helps why I'm so concerned at the moment is I'm working on This Project for my Homelab that I kind of need up and running yesterday, since the house DVRs and down and my girlfriend "needs" them. I intend to make both routers in the diagram Pf-Sense.
So inside of ZFS you can(or always?) make further nested layers of parity, like a Raid inside of the ZFS Pools? Is this done entirely in the command line or is there a utility that could perhaps hold my hand?
I had two drives die in a Raid 5 previously and lost it all, which has made me averse to single drive parity. This server is only going to be hosting VMs which isn't super duper important, but if I have an array go offline that takes VMs offline then suddenly my network is offline, which is a little problematic.
Well for ZFS from what I have read and also have heard on multiple tech shows is that if you are using zfs for vm storage you will want to use mirrors. So in your case maybe two mirrors only 1tb of space but if one dies than you cans till run and spend 40 bucks on a replacement drive and rebuild. You can also run a three deep zfs mirror also if you wanted to spend around 80 bucks and then that would give a bit more redundancy.
For simple virtualisation, I'm a big fan of Oracle VirtualBox. You can run it under Windows & Linux as the host OS and it is Open Source (GPLv2).
NB: You don't need pass-through for Ethernet access in VMs. Most VM implementations have a "bridged" mode, or a "NAT" mode which enables the Host's Ethernet connection to be shared with the VMs.
The difference is in "bridged" mode each guest OS needs a separate unique IP address (via DHCP or static) which is different from the host's IP address, and in NAT mode each guest uses the same IP address as the host OS via Network Address Translation (NAT) which is done by the virtualisation software.
I'm looking to do at least two Pf Sense VMs, I'm just concerned with extra latency being introduced if the VM doesn't have access to the "bare metal" as it were. Is that a likely issue?
Sounds like you're trying to defeat level 100, as a beginner. My advice is get aquainted with vbox on ubuntu server e.g. cli. learn how to script .sh scripts this will take a few, but .sh, and CLI will be your tools from now on, no more GUI, it's all about you taking control instead of M$ dictating what you can do, but once you're used to it, it's really a great tool. Once you feel you're ready to move onto level 2, try qemu, find a guide for hardware passthrough, and so on. combining iommu groups, ZFS, as a beginner really is a behemot of a task. Personally i've been working professionally with linux as a software engineer for several years, and when i tried iommu, and passthrough nvidia defeated me(They really are evil incarnate on the linux platform), and haven't even touched ZFS yet, my best is lvm pools.
See my link above, I need to best some NAT in order to have FiOS DVRs in the home work without having their router as my primary device. I know there's plenty of alternatives but the girlfriend loves old school setups.