SlackwareThis Forum is for the discussion of Slackware Linux.
Welcome to LinuxQuestions.org, a friendly and active Linux Community.
You are currently viewing LQ as a guest. By joining our community you will have the ability to post topics, receive our newsletter, use the advanced search, subscribe to threads and access many other special features. Registration is quick, simple and absolutely free. Join our community today!
Note that registered members see fewer ads, and ContentLink is completely disabled once you log in.
Originally posted by Fuel is it much better to have a separate /usr/local partition and install every app there .. what are the benefits ?
Note: its not a server its my personal everyday workstation .. games .. apps etc.
I think it comes in nicely when you decide to scratch your
base system and preserve /home and /usr/local with all the
time you invested to build it ... but then again, if you make a
major step in your glibc version your old stuff might not work
anymore, anyway... :} I guess it's a philosophical question ;)
The biggest disadvantage is that you'll have to be pretty
certain you know what space you're really going to need
for each partition ;)
If a bad thing happen to your Hd (your partition table disapears, heads crash, you type rm -rf by error...) you have better chance to save some data if you have made more than one partition.
Except that... no benefit.
I suppose every Linux system may vary to some extent, and I know the purpose can certainly vary a great extent. But my root file system consumes 2.7 GBytes, the /usr tree accounts for 2 GBytes of it. But the /usr/local tree is only 262 MBytes. Why even consider a separate partition for that very small percentage? (A whole 1% of the root file system.) And OpenOffice.org is 200 MBytes of that! A point was made of considering /usr as a partition, why I ask? At about 70% of the root file system, it *IS* the root partition.
I consider drives and partitions like unto baskets. Putting similar items together. These items are going to vary depending on the purpose of the server or workstation. A ftp server, mail server, and a web server, are all server configured machines but the data locations and needs may be different. A file server for a network would also be different. As is a user workstation. It makes sound sense to run quotas on a mail server, but would you run quotas on your own workstation? Since quotas apply to the partition where they are enabled, then it make sense that a mail server should highly consider /var/spool as a partition. A server that has a great number of users, then the /home tree should be highly considered for a partition, again because the probability of quotas is high and it is user data. How about a network file server, probably not using quotas, but you have shared data being stored and many times from a different OS. That data share should be considered a partition. Under these circumstances, the OS can be upgraded as needed without endangering the user data.
Enough of servers, what about workstations. What is the primary application? Production machine or just a normal type of user. Multiple users or single user? Some of both perhaps. These needs have to be considered. But again it comes down to data versus programs. Consider the basic root partiton as must for all systems. Then we should consider a partition for the user space, all be it /home or /root or where ever you decide the user space to be. The /home/user_name is not written in stone. It is specified in the passwd file when the user is added. One user can have the entire /home directory if desired. Or it can be /myfiles if you want. But regardless of the mount point it is a likely candidate for a partition. Now for production work. Well, graphics, audio, web, programming, etc. doesn't really make any difference except the amount of space required perhaps and maybe the number of partions, one, two or even three. For video and audio work, especially streaming, I would highly consider recommending separate drives!
If Linux software RAID is in use, then it operates on a drive/partition level very much like quotas. So that this may also be part of the equation when designing a server or a workstation. Consider; I have a server running with a 30 GByte and 20 GByte hard disk. These just happened to be the drives I had available at the time. I set up the server with a 10 GByte root partiton on the 30 GB drive. Then I setup the remaining 20 GB and the other drive as a RAID 1 (disk mirroring) partiton. If the 20 GB drive fails, replace drive and resync. If 30 GB drive fails replace, install OS, and resync. Either way the data volume is secure. That is good because it is too large to backup! Do you really want a server to backup that much everyday, over and over? It would spend more time doing backups than anything else.
The conclusion then is to let the application data decide. And not the OS or programs.
I know it has been a rather long post, but I like a good soap box every now and then. I rather like to think that it is a good reason for LQ being here. To share views. Perhaps someone else will share a different, even better, view. Thank you for reading and considering.
Great discussion Excalibur. Myself I tend to be a splitter. Part of this comes from the networked unix environment I learned in re: your discussion. Other factors involved: disks did not used to be so big, which meant without using multi-disk volume groups risking failure of all with the failure of one disk (disks used to be a little bit more expensive too (and less reliable), making redundancy hard to accomplish).
Despite these to some extent historical conditions, old habits are hard to break. My home system has about 10 scsi disks of varying sizes from 1GB up, and I just recently put on LVM, so certainly that has been a limiting factor. Also, I tend to have 3 or so distinct systems at anytime which share swap, /home, /tmp, /var, /pkg, /usr/src, and /usr/local partitions, but with separate (usually) /, /boot, and /usr partitions. Another reason I like separate partitions is for backup. I mirror my / and /boot partitions across the net to my work computer, using the -x with rsync makes this reasonable (and possible given my dialup connection) and much easier than other options I may have. Also, using -l with tar, it makes it easier for me to backup to tape what I consider logical units (is that why they call them logical volumes do you think?).
My general logic for file distribution is in part as follows:
/usr - generally the base distribution, rpm based
/usr/local - for programs I have built and installed myself
/pkg - the commercial software I use, most of which happens to need network derived licenses from my workplace.
Is this view better? No. It is just different and fits what I do, which is half or better of Excalibur's point. I decided to explain some of my logic and illogic just because it is. I thank the gods I use operating system within which I can design and manipulate setup to fit my needs rather than some other operating system I occasionally hear rumors of designed by people (in a windowless room :?) in a west coast city in N. America.
the conclusion then is to let the application data decide. And not the OS or programs.
one problem with redundancy...it protects you from hardware falures but it does not protect you from silly mistakes. Anything that happens go to both drives, so what your bork on one you will bork on the other. Taking this into account, maybe redundancy is a little overkill for a workstation...but a good incremental backup may save your arse.
since you have so much space, maybe boost the swap to 128 or 256, and I'd atleast go with a seperate /home partition, and whether you want a seperate /usr (mayb a seperate /var?) is up to you. I'd say just save most of the space for /home. Remember, you can also run personal bins (ie. /home/user/bin). A seperate /var is kinda useful if your logs go berzerk or something.
"my documents" replacement... when in windoze i have always hated "my documents"
/files is the place for my php projects, pictures, movies, mp3 etc and some backup. I only format / during upgrades/new install so i have to tweak, make sound work etc all over again wich is a good way to learn. When i have reached the point where i can handle the system a little bit better i will probably make some more partitions.
Back to the /usr/local .. lets se if i got this right..
./configure --help shows me the default install path and some misc info
./configure --prefix=/usr/local installs the app to /usr/local
but if you change de default install path what more do you have to do ?
maybe a separate /usr/local is only good for big games?
maybe make a partition named /games ?