Linux/UNIX Users Group at Michigan Tech
MTU LUG Logo Simple Linux Tux

File systems and storage

2021-11-18

Steven Whitaker

File systems and storage

introduction

Storage:

Memory:

Swap storage is fake memory that is actually storage, but pretends to be memory. You need a partition for this, and so you typemkswap /dev/sda2 to make the partition a swap file, then swapon /dev/sdXN to enable swap. Run top or htop to see your swap and memory usage.

This talk is all about storage, not memory.

Looking at things

lsblk "list block devices", shows partitions

blkid "block device ids", shows storage UUIDs

lsblk -f gives same info as blkid

fdisk -l list all your storage media

df -h shows you the free space on each drive.

Mounting systems

findmnt

Finds all the mounts in your system

Anything in /sys or /proc are kernel stuff. The kernel opens up files for you to look at. I don't know anything else about it! Look it up yourself, cuz I don't know what it means. It's not important to this talk.

/dev is your stuff... mostly. Here's some links to things I don't care about: /dev/pts. /dev/shm. /dev/mqueue.

Actually mounting shit

/dev/sdX are your drives! /dev/sda is your first, /dev/sdb is your second, etc.

You mount partitions.

mount /dev/sdb1 /mnt

umount /mnt

If you plug in a USB and you want to access it, use lsblk -f, then mount /dev/sdXN /mnt or wherever you want the USB files. Then, umount /mnt to safely remove.

(NOTE: if you're doing this on VirtualBox, your virtual drives have no format, so just run mkfs.ext4 /dev/sdX for every drive and you can mount them.)

Setting up the drives

fdisk /dev/sda

g sets disk as GUID Partition Table (GPT). This supports UEFI and up to 8 ZB (it goes GB, TB, PB, EB, ZB.. so a lot).

o Sets disk as a Master Boot Record MBR. This supports legacy boot, NO UEFI, and up to 2 TB size. Easier to set up than GPT.

n new partition

p primary partition (I have never used an extended partition).

w Writes what you've done

t Sets the file type. Key ones are: (MBR) 82=Linux swap, 83=Linux, fd=RAID, 8e=LVM. (GPT) 19=Linux swap, 20=Linux, 29=Linux RAID, 30=Linux LVM, 1=EFI. The default on fdisk is usually Linux.

Standard setup demo

Device        Start      End  Sectors  Size Type
/dev/sda1      2048   526335   524288  256M EFI System
/dev/sda2    526336  2623487  2097152    1G Linux swap
/dev/sda3   2623488 19400703 16777216    8G Linux filesystem

I would recommend 512M for the EFI system nowadays on Gentoo or systems that let you do this manually. Your /boot will get too clustered if you don't. I need to clean out /boot every 3rd kernel update and it can get annoying.

Logical volume management (LVM)

                   /---------------\ /-----\ /--------\
Logical volumes    | movies         | code  |  home   |
                   \---------------/ \-----/ \--------/

                   /----------------------------------\
Volume groups      |               lvm                |
                   \----------------------------------/

                   /----------\ /---------\ /----------\
Physical volumes   | /dev/sda1 | /dev/sdb1 | /dev/sdc1 |
                   \----------/ \---------/ \----------/

LVM pools multiple drives together in a single logical drive and let's you split them arbitrarily afterwards.

Device        Start      End  Sectors  Size Type
/dev/sda1      2048   526335   524288  256M EFI System
/dev/sda2    526336  2623487  2097152    1G Linux swap
/dev/sda3   2623488 19400703 16777216    8G Linux LVM
/dev/sdb1   2623488 19400703 16777216    8G Linux LVM
/dev/sdc1   2623488 19400703 16777216    8G Linux LVM

Format the volume: pvcreate /dev/sdX.

View the LVM settings

pvdisplay to view physical volumes, vgdisplay to view your volume groups. lvdisplay <volume_name> to view your partitions in the volume group. This stackoverflow answer helps explain the differences and gives two good links that helped me understand what was going on, technical and practical.

Main commands

Let us assume we have /dev/sda3, /dev/sdb1/, /dev/sdc1, 3 drives ready to be added to an LVM group. We will name our volume group "lvm" because why not? This will create a folder, /dev/lvm. First, add your drives to the physical volume list:

pvcreate /dev/sda3 /dev/sdb1 /dev/sdc1

Then create a volume group with them all:

vgcreate lvm /dev/sda3 /dev/sdb1 /dev/sdc1

Say you want your home directory in a different volume:

lvcreate -L 30G lvm -n lv_home

Then use the rest of the logical volume:

lvcreate -l 100%FREE lvm -n lv_root

You got a new hard drive? Cool. It's /dev/sdd1. Add it to the volume group:

vgextend lvm /dev/sdd1

You ran out of storage in a volume group? Resize it:

lvextend -L 35G /dev/lvm/lv_home

lvreduce to drop the size. This will probably break data.

RAID

RAID is for data protection. This is not a backup. It is just for protection. Instead of one drive dying and you lose all your data, you can use a RAID system so you have protection against it dying.

Instead of a table of the different RAID systems, here's a list of them. I'm going to assume we use RAID 1 or RAID 10.

RAID is done all under mdadm. It's straightforward comapred to LVM.

mdadm --create /dev/md0 --level=1 --raid-devices=2 /dev/sdb1 /dev/sdc1

mdadm --create /dev/md0 --level=10 --raid-devices=2 /dev/sdb1 /dev/sdc1

ZFS

You can use LVM and RAID by manually making sure you don't go over 50% or whatever RAID you want requires, then using mdadm on your logical volumes. Or just use ZFS to create a pool. Even less work.

zpool create pool /dev/sdb /dev/sdc

RAID 1:

zpool create pool mirror /dev/sdb /dev/sdc

RAID 10:

zpool create pool mirror /dev/sdb /dev/sdc mirror /dev/sdd /dev/sde

Check out what you did:

zpool status

Here is our Linux mirrors configuration:

  pool: lug
 state: ONLINE
config:

    NAME                       STATE     READ WRITE CKSUM
    lug                        ONLINE       0     0     0
      mirror-0                 ONLINE       0     0     0
        label/lug-HGST-4TB-01  ONLINE       0     0     0
        label/lug-HGST-4TB-02  ONLINE       0     0     0
      mirror-1                 ONLINE       0     0     0
        label/lug-HGST-4TB-03  ONLINE       0     0     0
        label/lug-HGST-4TB-04  ONLINE       0     0     0
      mirror-2                 ONLINE       0     0     0
        label/lug-HGST-4TB-05  ONLINE       0     0     0
        label/lug-HGST-4TB-06  ONLINE       0     0     0
      mirror-3                 ONLINE       0     0     0
        label/lug-HGST-4TB-07  ONLINE       0     0     0
        label/lug-HGST-4TB-08  ONLINE       0     0     0
      mirror-4                 ONLINE       0     0     0
        label/lug-HGST-4TB-09  ONLINE       0     0     0
        label/lug-HGST-4TB-10  ONLINE       0     0     0

errors: No known data errors

  pool: zroot
 state: ONLINE
  scan: scrub repaired 0B in 00:01:47 with 0 errors on Sun Sep 26 16:42:05 2021
config:

    NAME        STATE     READ WRITE CKSUM
    zroot       ONLINE       0     0     0
      mirror-0  ONLINE       0     0     0
        da10p3  ONLINE       0     0     0
        da11p3  ONLINE       0     0     0

errors: No known data errors

fstab

Automatically mount your file systems at start up and tells you which file system corresponds to which drive, etc.

<fs> <mountpoint> <type> <opts> <dump/pass>.

It is recommended you use UUIDs for non-LVM non-ZFS. It is recommended you use the device mapper symlink for LVM or ZFS. I am not smart enough to know the benefits for either.

Options are seen in "Filesystem independent mount options" and "Filesystem dependent mount options" in the mount(8) man page. I'd just use defaults. Use _netdev for a network drive (cough David cough).

The numbers at the end are not too necessary, but read the man page on the 5th and 6th fields if you really want to know. Gentoo recommends having / at 0 1 and everything else as 0 2.

```text /dev/lvm/lv_root / ext4 defaults 0 1 /dev/lvm/lv_home /home ext4 defaults 0 2 UUID=46ef4b37-b2bc-4a3a-b821-604c9fb64787 /boot ext4 defaults,noatime 0 2 UUID=61518d0a-1a81-4936-85e0-6bb7e1d69155 none swap sw 0 0