Linux Admininstration Resources?

shadfc asks: "I'm starting a new job as the system administrator for a small company in Tampa. They currently have 10 Red Hat servers (they are open to distribution change) that have not been actively maintained for a few months. I'm a Junior in College with a decent amount of Linux experience, but this will be my first job in this kind of position and responsibility. I'm asking for resources that can help fill in the holes in my knowledge and help make me a better administrator. Quality books on the subject would be preferred, but any advice is welcome. Thanks!"

Comments

Get a "playground" intranet box for experiments.

[DocSnyder]

Especially in enterprise environments, a wrong command or insufficient planning of some critical tasks can have severe side-effects. When I started administration, I installed GNU/Linux onto an old desktop PC which wasn't any longer good enough as a workstation but sufficient as a "playground" box. System upgrades, new kernel releases, complex shell scripts and even MTA or WWW server settings can be tested without disturbing other people's work. Internet access is only necessary as far as a HTTP proxy is concerned, to get updates.

Know your resources

[mnmn]

When I was learning Linux, I visited the Guides and HOWTOs every 5 minutes. www.linuxdoc.org and click on the sysadmin guide, networking guide etc.

To learn Linux itself, do a very basic install of a simple distro like slackware, or just a basic install of redhat on a test box, goto each directory like /etc, /sbin, /usr/sbin and read the man pages of every file you dont understand... for example you run into tune2fs, want to know what the heck is it, so you read the man page.

After a while you'll get the feel of Linux. You really dont have to know each command or how to use it.. man pages are available everywhere.

Try to compile your own kernel. That in itself teaches you alot about Linux and its capabilities. Beside that its the tools you have to know, such as apache, php, mysql, samba, nfs, ftpd, nmap, snort, sendmail/qmail/exim/postfix etc. Know the HOWTOs, guides, and man pages and youll never really need to buy books.

Any major problem you run into has already been fixed in the newsgroups. Goto groups.google.ca, and find your problem. Remember not to run Beta versions of services on your server for now... I'd even stay away from the 2.6 kernels until youve really tested the hardware on your side and are sure of it.

Re:You better patch them first

[nocomment]

Can you give us the IP addresses of these machines?

127.0.0.1/8


I'd recommend installing gentoo. It will take you some serious effort, but once you make it through it will seem easier the 2nd time. Install it about 4 or 5 times, and you will know a lot about linux. Then move on to another distro that doesn't take 3 days to compile the window manager like mandrake or debian (my personal favs in that order).

be warned however that you will get very frustrated if you try the gentoo thing. It will teach you how to manually create all your partitions, manually create your fstab, manually mount partitions, manually create and mount swap space, how to setup a chroot, etc etc... all of these things you will need to understand to administrate _any_ linux system, and many other unix variants. Have 2 systems running:1 that is on the gentoo install page (the docs are very good, read closely so you know when to skip ahead, i accidentally switched over to a stage 1 install because I skipped the section that said "if you are doing a stage3 skip to the next section), and the 2nd machine to actually do the install on.

ps. do the stage 3 install :-)

Why up2date?

[magefile]

If you're going to stick with RedHat see if you can get up2date working properly.

I'm not a sysadmin, I just use my home box (FC1, soon FC2), but in my experience, up2date is a slow, buggy, unreliable piece of crap. Go with yum. Not only is it faster and more stable, but you get more data from it, it allows you to install and uninstall stuff semi-automagically, and you can script it if you want.

Note: NEVER script upgrades on a production machine. Useful stuff to script would be "yum check-upgrade", and maybe add a file with a list of packages that you're anxiously awaiting an upgrade for (say, if you know that there's a security hole that will be fixed soon).

Re:You better patch them first

[dpilot]

Perhaps you miss *no comment*'s point. With most distributions, you can defer a lot of learning until later. Many times these days, later never comes.

With Gentoo, you really don't have the opportunity to defer a lot of learning. You need it just to get up and running.

Of course a diligent sysadmin *will* do the learning, but *no comment* advocates Gentoo as a not-so-gentle prod.

Re:Painful, but true.

[rusty0101]

Likewise, if you don't know when to start asking for help, you're probably the wrong guy as well.

Asking on slashdot is setting up a usable resource for finding many of the other resources you will probably never encounter otherwise. You could search Amazon, BN, and several Linux specific book resources, and never get an idea for how various books actually work out for the people who buy them. Ask on slashdot, and you will find out that author x in the second edition of book y, really couldn't find his ass if you told him to reach down and behind himself, then bring his hands forwards.

So far I see he has gotten referals to books that look like they are going to be great resources for pointing his users at when they have questions about the user side of the platform, a couple of administrator level guides to linux from O'Reilly, and several suggestions that he build a test box of his own to try out the things that he believes needs to be done on the servers he is becoming the administrator for.

I have also seen at least one recomendation to 'patch everything' which may, or may not be a good idea.

I have seen companies "patch everything" because they were using an outdated version of PHP, only to discover that the new version of PHP requires several completely different libraries, effectively taking the entire business offline for the week or two that it took to get all the dependencies resolved. If they had first tried the 'patch everything' approach on a dummy machine that was a mirror of the operating machine, they would have known what else needed to be done, and might not have had any business down time.

So, asking on slashdot suggests he is probably the wrong person? For a small shop, I don't think so. I think it shows that he is concerned, recognizes that there are several people here who do more than berate others, and allows him to pick out the good advice from the bad. Personally I would think the guy has a pretty good head on his neck.

But that's just my opinion, and I am not in the hiring department of the company that brought him in.

-Rusty

To start

[Beaker1]

http://www.google.com is your best friend

Nothing get's debuged on a production system. If it doesn't work it gets pulled off and fixed in the development environment.

Take root away from everybody and never give it out. Everyone has to learn this the hard way. Maybe you won't have to.

Standardize your OS installations and push back on mass customization. The users complain, but in the end they're more appreciative of a consistent working environment, then anything else.

Following these guidelines can help you sleep at night. When the pager goes off it's because a piece of hardware failed, not because some jackasses custom compiled perl installation that they didn't tell you about is chewing CPU and allowing hackers to use your systems as a pr0n site.