102.1 Lesson 1

Before a filesystem can be accessed on Linux it needs to be mounted. This means attaching the filesystem to a specific point in your system’s directory tree, called a mount point.

When mounted, the contents of the filesystem will be available under the mount point. For example, imagine you have a partition with your users' personal data (their home directories), containing the directories /john, /jack and /carol. When mounted under /home, the contents of those directories will be available under /home/john, /home/jack and /home/carol.

The mount point must exist before mounting the filesystem. You cannot mount a partition under /mnt/userdata if this directory does not exist. However if the directory does exist and contains files, those files will be unavailable until you unmount the filesystem. If you list the contents of the directory, you will see the files stored on the mounted filesystem, not the original contents of the directory.

Filesystems can be mounted anywhere you want. However, there are some good practices that should be followed to make system administration easier.

Traditionally, /mnt was the directory under which all external devices would be mounted and a number of pre-configured anchor points for common devices, like CD-ROM drives (/mnt/cdrom) and floppy disks (/mnt/floppy) existed under it.

This has been superseded by /media, which is now the default mount point for any user-removable media (e.g. external disks, USB flash drives, memory card readers, optical disks, etc.) connected to the system.

On most modern Linux distributions and desktop environments, removable devices are automatically mounted under /media/USER/LABEL when connected to the system, where USER is the username and LABEL is the device label. For example, a USB flash drive with the label FlashDrive connected by the user john would be mounted under /media/john/FlashDrive/. The way this is handled is different depending on the desktop environment.

That being said, whenever you need to manually mount a filesystem, it is good practice to mount it under /mnt. The specific commands to control the mounting and unmounting of filesystems under Linux will be discussed in another lesson.

On Linux, there are some directories that you should consider keeping on separate partitions. There are many reasons for this: for example, by keeping bootloader-related files (stored on /boot) on a boot partition, you ensure your system will still be able to boot in case of a crash on the root filesystem.

Keeping user’s personal directories (under /home) on a separate partition makes it easier to reinstall the system without the risk of accidentally touching user data. Keeping data related to a web or database server (usually under /var) on a separate partition (or even a separate disk) makes system administration easier should you need to add more disk space for those use cases.

There may even be performance reasons to keep certain directories on separate partitions. You may want to keep the root filesystem (/) on a speedy SSD unit, and bigger directories like /home and /var on slower hard disks which offer much more space for a fraction of the cost.

The swap partition is used to swap memory pages from RAM to disk as needed. This partition needs to be of a specific type, and set-up with a proper utility called mkswap before it can be used.

The swap partition cannot be mounted like the others, meaning that you cannot access it like a normal directory and peek at its contents.

A system can have multiple swap partitions (though this is uncommon) and Linux also supports the use of swap files instead of partitions, which can be useful to quickly increase swap space when needed.

The size of the swap partition is a contentious issue. The old rule from the early days of Linux (“twice the amount of RAM”) may not apply anymore depending on how the system is being used and the amount of physical RAM installed.

On the documentation for Red Hat Enterprise Linux 7, Red Hat recommends the following:

Of course the amount of swap can be workload dependent. If the machine is running a critical service, such as a database, web or SAP server, it is wise to check the documentation for these services (or your software vendor) for a recommendation.

We have already discussed how disks are organized into one or more partitions, with each partition containing a filesystem which describes how files and associated metadata are stored. One of the downsides of partitioning is that the system administrator has to decide beforehand how the available disk space on a device will be distributed. This can present some challenges later, if a partition requires more space than originally planned. Of course partitions can be resized, but this may not be possible if, for example, there is no free space on the disk.

Logical Volume Management (LVM) is a form of storage virtualization that offers system administrators a more flexible approach to managing disk space than traditional partitioning. The goal of LVM is to facilitate managing the storage needs of your end users. The basic unit is the Physical Volume (PV), which is a block device on your system like a disk partition or a RAID array.

PVs are grouped into Volume Groups (VG) which abstract the underlying devices and are seen as a single logical device, with the combined storage capacity of the component PVs.

Each volume in a Volume Group is subdivided into fixed-sized pieces called extents. Extents on a PV are called Physical Extents (PE), while those on a Logical Volume are Logical Extents (LE). Generally, each Logical Extent is mapped to a Physical Extent, but this can change if features like disk mirroring are used.

Volume Groups can be subdivided into Logical Volumes (LVs), which functionally work in a similar way to partitions but with more flexibility.

The size of a Logical Volume, as specified during its creation, is in fact defined by the size of the physical extents (4 MB by default) multiplied by the number of extents on the volume. From this it is easy to understand that to grow a Logical Volume, for example, all that the system administrator has to do is add more extents from the pool available in the Volume Group. Likewise, extents can be removed to shrink the LV.

After a Logical Volume is created it is seen by the operating system as a normal block device. A device will be created in /dev, named as /dev/VGNAME/LVNAME, where VGNAME is the name of the Volume Group, and LVNAME is the name of the Logical Volume.

These devices can be formatted with a desired filesystem using standard utilities (like mkfs.ext4, for example) and mounted using the usual methods, either manually with the mount command or automatically by adding them to the /etc/fstab file.