Sooner or later it happens: you type "df" and see that the disk is pushing 99%. Or, worse, you get awful messages on the console about no space.
(If you are out of space right now, follow this link for suggestions as to what to do right now).
So, it's time to add more disk space. There are two basic ways to approach this problem. You could replace the current disk with something larger, or simply add another disk. Both methods have their own positive and negative aspects. Let's consider some of them:
Replace the current disk
This can be the simplest solution of all, particularly if you are using emergency boot and recovery disks such as can be made by Microlite Edge or other similar products. Do a backup, pop out the old disk, in with the new, boot with the Recovery disks, and very quickly you have a running system.
But this assumes that you either have some other use for your current disk, or that it's old and worthless enough that you are willing to toss it out. That may be the case, but in those situations, you may have a further problem if you have to replace the old disk with a different architecture. You may have an old IDE disk, and want to replace it with a new SCSI Raid system, for example. In those situations, the Emergency Recovery disks may not be enough, and you may have to start with a fresh install of the OS and proceed from there.
Add another disk
Adding storage can, of course, involve similar trade-offs. You could add a SCSI drive to an existing IDE drive, for example, but generally not vice-versa. You also need to decide what information you are going to store on that new space, and how you are going to get it there. Some programs have the ability to make use of new space fairly simply, but more often than not, you are going to have to do some data rearranging.
See Adding a disk drive for details on how to add the disk
Dividing the disk
Whichever way you decide to go, you need to start thinking about partitions before you do anything. Partitions, as you may now, are the areas that your hard drive is divided into.
It can be a little confusing on SCO, because there are really two kinds of divisions: "fdisk" partitions and "divvy" partitions (which are called "divisions" by SCO, but "partitions" by some other Unixes, which is why it gets confusing).
If you have an older version of SCO, you may have to be concerned about fdisk partitions for very large drives. More modern versions are able to support about anything you could conceivably purchase (assuming proper drivers, of course). If you have SCO 3.2v4.2, you will probably need or want uod429a for EIDE or large SCSI drives. This supplement has some confusing aspects to it, so read the README file carefully and completely.
I'll assume for the moment that you have no issues that cannot be resolved by proper drivers or supplements, and that you will be creating one Unix/Linux fdisk partition. It's now time to think about divvying up the disk.
You'll be dividing the Unix partition into filesystems or raw disk partitions. Raw partitions are strictly for the use of programs like Informix that are written to use these partitions, generally to obtain higher performance. Without such programs, your divisions will be used for ordinary Unix filesystems. Older systems usually had just a root division or possibly a root and a /u (The "root" division is the one that shows up as "/" mounted on "/dev/root").
(SCO Release 5 systems usually have a root and a small /stand. Linux may have a "/boot").
You need to decide what partitions you will create, and how large they will be.
On older releases, the OS constrains your choice to some degree because it may not be able to handle a filesystem larger than 2 gigabytes. That was the case with SCO 3.2v4.2: If you had a raw partition, that could be larger, but any filesystem prior to 3.2v5.x had to be 2 gigabytes or less.
Why divide at all?
Assuming Linux, SCO Release 5, or a 4.2 system with a 2 gig drive, why not dispense with extra divisions and simply have one large file system?
You could do that. With release 5, you'll probably still want a small (15-20MB) /stand filesystem, because Unix can't boot from an HTFS filesystem, and that's what you would want the root to be. The /stand would be an (older) EAFS filesystem, which Unix can boot from, and the root would be the higher performance and more reliable HTFS type.
If you are planning to use LVM with Linux, you may also want a "normal" /boot filesystem.
See Partitioning for Linux also.
Even on old SCO 3.2v4.2, you could have one large root filesystem (up to 2 gigabytes). It's very convenient to do this; you don't have to worry about how much space to allocate to what, you just let the whole disk be one big filesystem.
For several reasons, you probably don't want to do that. On the older releases (and even on the newer, under some conditions), big file systems can take a long time to clean and repair in the event of an unexpected crash. It is also true that damage to a hard drive usually confines itself to one file system. In other words, if a defective hard drive trashes your root filesystem, a separate /u file system may escape damage entirely. This can make recovery simpler and more complete, because the data on /u may be current right up to the crash.
Deliberately limiting space
You may also want to create file system divisions to specifically limit how much space can be used. At first, this may seem silly, but consider the often seen case where a naive user keeps submitting the same large print job over and over, not understanding that it is not printing for some other reason (printer offline?) and that each submission is using up more disk space. Consider also a runaway program that keeps creating a larger and larger file. Wouldn't it be nice to let a separate file system run out of space before the whole drive is used up? In the case of the printing, you could create a small division and mount it at /var/spool/lp/temp.
Note that you have to be a little bit careful about creating partitions that will be used for system directories, like /user, and on Release 5, the mess of symbolic links means you have to tread even more carefully. With enough care and forethought, you can do whatever you need to do. You have to watch out for things that need to be present when the system is booting; if you are planning on moving anything like that elsewhere, you will have to leave some of it behind, ultimately to be hidden beneath your new file system. This can get tricky, so be careful or hire competent help.
Deciding what to move
You may have one large area that will continue to grow. For example, /usr/mas90, /appl/filepro, and /usr/rwc65 are common SCO programs that have defined directory structures. If you have or can identify such an area, then this is the data you will move.
You may, however, have two or more areas to move. You can solve this by creating multiple file systems in the new space, or by using one file system and symbolic links to point to the new data areas.
There are things you cannot move. You cannot, for example, move /usr or any part of it that needs to be present as the machine boots or in single user mode. DO NOT MOVE ANYTHING THAT ISN'T YOURS and you'll be safe.
If you are installing a replacement drive, you are not concerned with moving data. If, for example, your old system had a large /data filesystem, you could simply specify a larger division when divvy'ing the new disk. The restore will simply put the old data in the new, larger, division.
But when adding a new hard drive, you may want to move the contents of a directory to the new space. The following procedure is useful:
Create the new space with "mkdev hd" as required. Relink, and reboot. Rename the directory you are going to move, using the "mv" command. For example:
mv /usr/data /usr/data.safe
Note the ownership and permissions of /usr/data.safe, and create a new /usr/data directory with the same permissions and ownership. Example:
# cd /usr # mv data data.safe # ls -ld data.safe drwxr-xr-x 2 root sys 512 Nov 16 18:47 data.safe # mkdir data # chown root data # chgrp sys data # chmod 755 data # ls -ld data drwxr-xr-x 2 root sys 512 Nov 16 18:47 data
Run "mkdev fs", giving the name of the directory you want to move as the mount point (/usr/data). Then type:
mountall to mount the new space at /usr/data (or
simply type "mount /usr/data").
Type the following to copy the data:
# cd /usr/data.safe # find . -depth -print | cpio -pdlmv ../data
After this is done, and you are completely satisfied that everything is working as it should be, you can "rm -r /usr/data.safe"
If you are using symbolic links, the procedure is similar. First, create a new directory on the other drive:
Assign appropriate permission to
Copy the data as before, and mv the oldname aside:
cd /usr/data find . -depth -print | cpio -pdlmv /bigdrive/newdata mv /usr/data /usr/data.sf
Now create the symbolic link:
ln -s /bigdrive/newdata /usr/data
Again, when you are satisfied that everything is working, come back and "rm -r /usr/data.safe".
Got something to add? Send me email.
More Articles by Tony Lawrence © 2009-11-07 Tony Lawrence
That's the thing about people who think they hate computers. What they really hate is lousy programmers. (Larry Niven)