[Background]
I've been a long time dabbler with Gentoo whilst I kept Winders around on all of my desktops for a variety of reasons, usually installing Gentoo in either VMs or else on separate drives and manually controlling the boot drive for my system via BIOS / UEFI.
This year marks a change for me in that the rig I spec'd and built in 2020 in anticipation for Winders11, and used successfully for this entire time, until last month. Finally fed up with the idiosyncrasies of Micro$oft, and my trials and tribulations with trying to fix this or that broken thing, coupled with the fact that my current job now (for well over 2 years) is 99% web based and is not tied to any single OS ecosystem at all, made me realize it was finally time to break from this addiction to M$.
That being said, I have a few questions, some being just a sanity check, and some being more ... esoteric.
First off, my hardware:
- MSI MEG X570 Godlike mobo
- Ryzen 9 3950X
- 128 GB RAM
- RTX 3080 Ti
- 3x Samsung 980 Pro 2 TB NVMe drives
Now, I've been playing around with a few distros for various reasons over the last couple of weeks because I needed some things up and available to me ASAP, but I do very little coding in my role, and outside of work I *do* game, a lot, mostly Android and Steam. I'm not thrilled with the idea of Bazzite being a locked down system (though you can enable root), Nobara seems OK, CachyOS seems limited, and, anyway, I've always had a soft sport for Gentoo over the decades.
I've done an intense amount of reading, considering, and fact-checking on various things regarding a bare metal install on this machine, and I've decided that I want to make use of all 3 NVMe drives as well as the RAM and make this a machine that I can build out the way that I want. What I *don't* want is keeping hte drives independent, with dedicated partitions, and neither do I want to use LVM - been there, done that in the past, just not my cup of tee.
I'm looking at btrfs instead, for a few reasons, including it's ability to pool drives, use subvolumes (and nested subvolumes), and more, as well as maybe throw in some tmpfs magic for compiling, and generally set up install so I have the 5+ TB of space readily available to me through the magic (or shenanigans, depending upon how you view it) of btrfs. All three drives were purchased and installed at the same time, replacing 3x Sabrent Rocket4 drives of half the capacity, which now are accessible via a USB3 Hub to the system with backups and everything I could ever need if I ever *have* to have Winders installed again.
However, therein lies my dilemma. I know that I can stripe across all three after creating my obligatory ESP, swap, and boot (? I didn't think this was necessary, but I saw at least one person doing it in the forums) partitions, but as I want to game, and reading so much material that btrfs is not necessarily recommended for large game files, among other tons of tidbits of info, here come the questions.
[Questions]
1) Should I use all 3 drives under one fs, or should I split at least one out as a dedicated gaming part using e.g. ext4 as its fs?
2) Should I use tmpfs for compiling? Background: one of my drives has 7% life expectancy used (formerly the main OS drive for Winders, as it got reinstalled many, many times over these few short years), whereas the other 2 are both at 3%. All of this is still within acceptable limits, as I bought the drives over 4 years ago now, and even the one with more wear / use should still easily last me another 15 years if I were to actually keep using it that long.
3) Is it worth considering disabling compression on the btrfs pool? I searched far and wide for benchmarks on btrfs with compression disabled, and have not found anything really reliable, but I found plenty of instructions on how to run btrfs without compression (as well as the fact the compression levels run in a range from -15 to 15 for zstd). But I also found seemingly contradictory information that the compression actually helps with the read speed of files (I'm assuming these were from platters and / or really slow SSDs).
4) Assuming I'll be both compiling programs from source as well as installing binaries here and there, is it worth looking at an aggressive set of subvolumes to really section off various locations from each other?
In the meantime I'm going to do a smaller single disk install to make sure I 'can read the handbook correctly' and get a test system up and working.
Any thoughts, suggestions, and even criticisms are welcome. Thanks.
