Differences
This shows you the differences between two versions of the page.
| Both sides previous revisionPrevious revisionNext revision | Previous revision | ||
| selfhost:storage [2024/10/24 12:29] – willy | selfhost:storage [Unknown date] (current) – removed - external edit (Unknown date) 127.0.0.1 | ||
|---|---|---|---|
| Line 1: | Line 1: | ||
| - | ===== Storage Setup ===== | ||
| - | |||
| - | Your home server | ||
| - | During the Gentoo [[gentoo: | ||
| - | |||
| - | You have already installed your Gentoo Linux, fresh and nice. Now it's time to organize and setup the storage. | ||
| - | |||
| - | First of all: the disk/ | ||
| - | |||
| - | The best approach is to use one drive for the OS and different drives (usually at least 2) for your data and services. These drives should be physically separated drives, and not different partitions on the same drive. | ||
| - | |||
| - | So you will have your **OS** drive and your **services** drive (which will include also the data). | ||
| - | //note:// having data and services on the same drive ensures that you can perform hard-links, which in some cases are mandatory to avoid data duplication (i am talking about torrents). | ||
| - | |||
| - | The idea is to store your **services** on a redundant software RAID-1 array. Now, there are different solutions you can choose from. You could go RAID-5, RAID-0+1, RAID-6, RAID-10 and many more combinations. I will let you research and find out what works best for your use case and drives availability. | ||
| - | |||
| - | RAID-1 has a few advantages for me which are: | ||
| - | * Fast enough on read (reads will be balanced on both disks, writes will take slightly longer) | ||
| - | * Solid enough to survive one disk fail (provided you monitor the RAID status and replace failed disks) | ||
| - | |||
| - | It's main drawback is the wasted space: 50% is a big waste, but it's for the peace of mind. Just remember: **RAID does NOT means BACKUP** and **do your backups** anyway. More on this later on. | ||
| - | |||
| - | (note: i will always say //disk// but that can be an SSD, doesn' | ||
| - | |||
| - | ==== Software or Hardware RAID? ==== | ||
| - | |||
| - | RAID can be implemented in hardware or in software, using Linux RAID software implementation. | ||
| - | I have been using the software RAID approach for more than two decades and i have never been let down: | ||
| - | * It's solid, | ||
| - | * it's simple, | ||
| - | * it works and it's efficient. | ||
| - | * Each disk can always be mounted as single drive, without the RAID array. | ||
| - | |||
| - | If you choose to use a commercial external RAID solution, skip the RAID part ahead. | ||
| - | |||
| - | ==== Disks Preparations ==== | ||
| - | |||
| - | I will assume you have two external drives called **/ | ||
| - | |||
| - | The size of the two disks is not important: get the biggest ones you can afford. | ||
| - | |||
| - | I suggest, if you can afford them, to use SSDs because more silent and consume less power, which is a plus for a home server, but they are still more expensive than traditional drives. | ||
| - | Any way, it doesn' | ||
| - | |||
| - | A good approach to add more drives, when you run out of internal slots in your server, is to use USB-3/USB-C external drives. You can buy a JBOD box (Just a Bunch Of Disks) where you can store 2, 4 or even 8 or 16 disks sharing one USB plug. I have been using this type of setup for the best of the last 15 years without any data loss or corruption. Speed-wise, you will be streaming your data over your home network, which more than often means WiFi. A good USB-3 SSD is more than capable to keep up any data transfer requirement for any streamed media today, even 4K, so there is not need to worry that external disks or USB-3 might be a bottleneck. | ||
| - | |||
| - | Note: i will refer to //two// disks, but you can create more RAID arrays if you have //four// disks! | ||
| - | |||
| - | ===== Partitioning ===== | ||
| - | |||
| - | To create a software RAID, you need to first partition the two drives, for this job you can use the good old fdisk: | ||
| - | |||
| - | <code bash> | ||
| - | su | ||
| - | fdisk /dev/sdb | ||
| - | ... do the partitioning ... | ||
| - | fdisk /dev/sdc | ||
| - | ... do the partitioning ... | ||
| - | </ | ||
| - | You will need to be root for fdisk to work. You should be root at this point, tough. The //su// command might be redundant. How to use fdisk? I think you can find out easily. given that these should be new and clean disks, there is not much risk to mess up. Create a GPT partition table, for future-proof support, and one single partition filling up the disk, unless you want a more complex setup. | ||
| - | |||
| - | Using //fdisk//, create one partition on each drive to fill it, that will be called **/ | ||
| - | |||
| - | Save the changes and quit from fdisk, since the disks are not being used yet, you will not need to reboot the server. | ||
| - | |||
| - | Remember that using Linux Software RAID you can create more than one partition and create more than one RAID-1 from two disks. For example, if you have a huge disk and want to separate two areas (one for data and one for webcam storage, for example) you can create **two** RAID-1 arrays by splitting both disks in two partitions each and mixing them up. Just don't create a RAID from two partitions on the **same** disk as that would be, at best, dumb. | ||
| - | |||
| - | (if you need to retain your data and you have only two disks, you can create the RAID only on one of the two, which will be deleted, mount with only one drive, copy the data over from the other disk, then format the other disk and hot-add it to the RAID-1. How to do this in details it's not difficult to figure out, but be careful not to lose your data in the process) | ||
| - | |||
| - | |||
| - | ===== Creating the RAID array ===== | ||
| - | |||
| - | You need to create a new software RAID array out of **/ | ||
| - | |||
| - | <code bash> | ||
| - | mdadm --create /dev/md0 --level=1 --raid-devices=2 /dev/sdb1 /dev/sdc1 | ||
| - | </ | ||
| - | |||
| - | since this is the first RAID array of the server (and probably the only one) it's called **/ | ||
| - | |||
| - | If you want to do the trick of copying existing data from one of the two disks, at this point you can create a RAID-1 array with only one drive by replacing the drive you do not want to add now with the work // | ||
| - | |||
| - | |||
| - | ===== Format and mount ===== | ||
| - | |||
| - | You need to format and mount your newly created RAID array. You need to choose which filesystem you want to use. A common choice on Linux, and probably the more straightforward, | ||
| - | |||
| - | Format the RAID array: | ||
| - | <code bash> | ||
| - | mkfs.ext4 /dev/md0 | ||
| - | mkdir /data | ||
| - | mount /dev/md0 /data | ||
| - | </ | ||
| - | |||
| - | The newly formatted drive needs to be automatically mounted at every boot, so you need to add a line like this to **/ | ||
| - | < | ||
| - | / | ||
| - | </ | ||
| - | |||
| - | The //noatime// option will reduce USB traffic and wear-and-tear. Of course, change the filesystem type to whatever you choose. | ||
| - | |||
| - | ===== Automate RAID at boot ===== | ||
| - | |||
| - | You also want to automate linux raid startup, so that upon reboot everything will still work just fine. To do so, the **mdraid** service needs to be started in the //boot// runlevel. Do NOT start it in the //default// runlevel or things will break badly after the first reboot. | ||
| - | |||
| - | <code bash> | ||
| - | > rc-update add mdraid boot | ||
| - | </ | ||
| - | |||
| - | The //mdadm// service is not required, unless you want monitoring of your RAID array (nicluding email reporting) which is a **TODO**. | ||
| - | |||
| - | You also want to ensure the **/ | ||
| - | |||
| - | < | ||
| - | ARRAY /dev/md0 UUID=1758bcfa: | ||
| - | </ | ||
| - | |||
| - | where the UUID can be read by the output of the command: | ||
| - | |||
| - | <code bash> | ||
| - | mdadm -detail /dev/md0 | ||
| - | </ | ||
| - | |||
| - | One last bit that might be required if you use USB storage is to increase the udev timeout at boot. USB drives might be slow to spin-up and be recognized (even SSDs) so this trick might be needed if you find that upon reboot your RAID array has not been mounted properly. | ||
| - | |||
| - | Add this lines to your **/ | ||
| - | < | ||
| - | start() { | ||
| - | local output | ||
| - | ebegin " | ||
| - | sleep 10 # line added | ||
| - | ebegin " | ||
| - | </ | ||
| - | Adapt the ten seconds to your likings. | ||