- Chapter 3 Linux Filesystem
- The Linux Filesystem
- Filesystem Organization
- Keeping Your Disks Healthy
- Summary
Keeping Your Disks Healthy
After your partitions are created and formatted and you have installed Linux, you will want to make sure they stay healthy. You can use several different utilities to determine the state of your filesystem or fix problems with it.
fsck
The most important utility for maintaining your filesystem is the fsck command. Like the mkfs command, fsck is a front end to the check disk command for the specified filesystem type. It is used to verify the filesystem's structure and repair any inconsistencies. Problems most frequently occur as a result of a system crash.
When a system crashes, the kernel is unable to sync the cache with the hard disk's contents. This can result in blocks being marked as in use when they are really empty. Other errors can occur from directly writing to a device to an area containing existing data.
The syntax for the fsck command is
fsck -t fs-type device
For example, to check the first partition created as an ext2 filesystem in our earlier example, you would type the following:
fsck -t ext2 /dev/hda1
The previous command calls the e2fsck command to check the ext2 filesystem on the first partition of the first IDE hard disk. Because this is the root partition, you will receive a message that the filesystem is mounted and be asked whether to continue. If you answer yes, fsck checks the following:
inodes
blocks
file sizes
directory structure
links
Any partition other than root can be unmounted while your system is running. To run the fsck command, you must first unmount that filesystem. See Chapter 6 for more information on mounting and unmounting filesystems. To run fsck on your root partition, you can
Start in read-only, single-user mode.
Boot with a floppy and then scan the root partition.
Run fsck during system boot.
When running fsck during system boot, the root filesystem is mounted in read-only mode, fsck runs, and then the root filesystem is mounted as read-write. All other filesystems are checked before they are mounted. Refer to Chapter 2, "Booting Linux," for details on booting and passing parameters to your boot loader.
Key Concept
The fsck command is used to verify the integrity of your filesystem. Unmount your filesystem before running it.
Options that are filesystem-specific for the fsck command are listed in Table 3.7.
Table 3.7 The Filesystem-Specific fsck Options and Their Actions
|
Option |
Action |
|
-a |
Causes fsck to run non-interactively |
|
-c |
Checks for bad blocks |
|
-v |
Reports on fsck's progress |
The filesystem keeps information about the filesystem as a whole in the superblock. If the superblock becomes corrupted, the filesystem cannot be mounted; however, backup copies of it are spaced at regular intervals across the filesystem. By default, these copies are saved every 8,192 blocks. Because the first block is number 1 and not number 0, copies would be stored at block 8193, 16385, 24577, and so on. To verify that this is the block group size on your system, you can issue the following command:
dumpe2fs device
The previous command produces a wealth of information about the filesystem; just look for the following line:
Blocks per group: 8192
If a different number is given, use this number to locate the copies of the superblock. You can then tell the e2fsck command to use a copy of the superblock to check a partition you can't mount because of a corrupted superblock. So, if you can't mount a /dev/hda1 that has an ext2 filesystem because of a corrupted superblock, issue the following:
e2fsck -f -b 16385 /dev/hda1
In the previous command, the -f option is used to force the check. If you do not use the -f option, e2fsck will look at the copy of the superblock and might think that the filesystem is clean. The -b option identifies the copy of the superblock that is to be used.
Disk Usage
As files are created and saved to your system and log files are written to, your free disk space shrinks. It is important to monitor how much space is used on each disk. Linux runs best when there is adequate free space available.
To get optimal performance from your operating system, the amount of free space available should be 5–30 percent of each filesystem. In addition, if there is no free space, you will not be able to write to that filesystem.
Not only do you need to monitor the amount of free space, but you also will want to monitor what files are occupying the most space. By seeing what files are growing the fastest, or who is using the most space, you will be able to take steps to prevent your filesystem from becoming too full.
du
The du command reports on the amount of space occupied by a single file or a directory and all its files and subdirectories. The syntax for the du command is as follows:
du [options] [filename]
In the previous syntax, filename can also be the name of a directory. If du is called without a filename, it will report on the working directory and its contents. Table 3.8 lists the options you can use with du.
Table 3.8 The Options Used with the du Command
|
Option |
Action |
|
-a |
Displays file usage |
|
-b |
Displays usage in bytes |
|
-c |
Prints grand total |
|
-h |
Appends a letter to identify the measurement being used for the report, such as M for megabytes |
|
-k |
Displays usage in kilobytes, the default |
|
-m |
Displays usage in megabytes |
|
-l |
Counts links |
|
-s |
Prints only the grand total |
Unless you use the -a option, du displays only the directories, even though the space occupied by all the files is counted. If you do not have permission to a file, du will display an error message and not count that file when displaying disk usage.
df
Both the du and df commands report on object size. Whereas du reports on the size of objects such as files, df reports on the size of devices such as partitions. The syntax for the df command is
df [options] [filename]
The df command returns the following information values:
Size of the device
Number of free blocks on the device
Number of occupied blocks on the device
Percent of total blocks that are free
Name of the device
If you pass an ordinary file's name to df, it will report on the device containing that file. If you do not name a file, df reports on all mounted filesystems. The options used with the df command are listed in Table 3.9.
Table 3.9 Options Used with the df Command
|
Option |
Action |
|
-a |
Displays information on all filesystems; default action |
|
-h |
Appends a letter to identify the measurement being used for the report, such as M for megabytes |
|
-i |
Lists inode usage |
|
-k |
Displays usage in kilobytes |
|
-m |
Displays usage in megabytes |
|
--sync |
Calls the sync command before gathering information |
|
-t fs-type |
Displays only filesystems of fs-type |
|
-T |
Displays the filesystem type for each entry |
|
-x fs-type |
Excludes filesystems of fs-type |
For example, if you type
df
your output might look like the following:
Filesystem 1024-blocks Used Available Capacity Mounted on /dev/hda1 1018298 93756 871931 10% / /dev/hda9 2555450 26494 2396823 1% /home /dev/hda7 132185 31 125328 0% /tmp /dev/hda5 2035606 468660 1461722 24% /usr /dev/hda6 256592 9435 233905 4% /var
Key Concept
Use the df and du commands to determine how much free space is available.
What do you do if a filesystem is becoming too full? You will want to identify any large files that are being created but are no longer necessary. Log, mail, and news files are often big offenders. Other large files that might be deleted are core dumps.
Check whether the temp directories are not being cleaned out. In fact, this should be a regular part of your system maintenance.
Are there files that are not being used? Try compressing them and perhaps relocating them to another device. Directories also should not get too large because very large directories can affect system performance. Move files to several smaller directories.
One of the problems with directories is that they grow but do not shrink. So, if you move or delete a large number of files from a directory, you will not recover all the free space.
The only way to shrink the size of a directory is by deleting it. Therefore, copy the files you want to keep to another directory and then delete the original directory. After you've done this, you can re-create the directory with the same name and move the needed files to the new directory.
If you are outgrowing a partition, you have a couple choices for fixing the problem.
One possibility is to move some files to another partition and then place links to them on the original partition. See Chapter 6 for information on how to use links.
A better way would be to actually increase the size of the partition. However, this is also a more complex process. The first thing you must do is to make sure you have an up-to-date backup of the information on that partition. See Chapter 11, "Backup and Restore," for information on performing backups.
Next, you must decide where the extra space will come from. You can use any free space that exists on the same hard disk or free space from another hard drive, or you can add another hard disk.
You must delete the old partition and then create a new partition to take its place. Partitions can't be dynamically expanded under Linux; they must be deleted and re- created with a larger size. After you have created a larger partition, you will need to restore the original information. The restore operation is also covered in Chapter 11.