Storage management

From MTULUG
Jump to:navigation Jump to:search

Introduction

Save large files like code, binary files, movies, etc. The storage media are things like HDDs, SSDs, magnetic tape.

Storage is not memory, where memory caches memory for quick access, things like any type of RAM: SRAM, SDRAM, VRAM, etc.

Swap storage is fake memory that is actually storage, but pretends to be memory. You need a partition for this, and so you run

root #typemkswap /dev/sda2

to make the partition a swap file, then run

root #swapon /dev/sda2

to enable swap. Run top or htop to see your swap and memory usage.

Looking at things

  • lsblk - "list block devices", shows partitions
    • lsblk -f gives same info as blkid
  • blkid - "block device ids", shows storage UUIDs
  • fdisk - list all your storage media using fdisk
  • df -h - shows you the free space on each drive.
Tip
When trying to save space, use du -sh ./* to look at folder sizes and work your way down to delete the things that aren't needed.

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 working with storage systems.

/dev is your stuff... mostly. Some system stuff that you don't use: /dev/pts, /dev/shm, /dev/mqueue.

Mounting commands

You're working with storage media, so every single command will always be run as root.

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

Mount a drive to look at:

root #mount /dev/sdb1 /mnt

Unmount it when you're done:

root #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 testing this on VirtualBox, your virtual drives have no format, so just run
root #mkfs.ext4 /dev/sdX
for every drive and you can mount them.

Bare-metal drives

We use fdisk.

root #fdisk /dev/sda
fdisk commands
Command What it does
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.
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.
w Write your changes.

Demo setup

Follow this setup to test.

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
Tip
I'd recommend playing with this in VirtualBox because it is very difficult to change these on your filesystem once you have data on it. In fact,
Warning
You destroy all data when running these commands, so if you're doing this on your system just move everything off the drives before running these commands.

Logical volume management (LVM)

LVMs abstract away drives quite a bit.

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

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

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

The main commands are:

LVM commands
Command What it does
pvcreate Add a drive to the physical volumes.
pvdisplay View physical volumes in the LVM.
vgcreate Create a logical group with specified physical volumes.
vgextend Add physical volumes to the volume group.
vgdisplay View logical groups in the LVM.
lvcreate Create a logical volume within the volume group.
lvextend Expand a logical volume within the volume group.
lvreduce Decrease a logical volume within the volume group.
lvdisplay View all logical volumes.

I'd highly, highly recommend following along with this tutorial within an LVM. You will be very comfortable with all of these commands after this.

Warning
You destroy all data when running these commands, so if you're doing this on your system just move everything off the drives before running these commands.

Demo setup

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 that contains our logical volumes.

First, add your drives to the physical volume list:

root #pvcreate /dev/sda3 /dev/sdb1 /dev/sdc1

Then create a volume group with them all:

root #vgcreate lvm /dev/sda3 /dev/sdb1 /dev/sdc1

Say you want your home directory in a different volume:

root #lvcreate -L 30G lvm -n lv_home

And then use the rest of the logical volume:

root #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:

root #vgextend lvm /dev/sdd1

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

root #lvextend -L 35G /dev/lvm/lv_home

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.

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

root #zpool create pool /dev/sdb /dev/sdc

RAID 1:

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

RAID 10:

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

Check out what you did:

root #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.

Here's my computer's setup:

/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