One of our users, James Lowden, emailed us to say that our recent 77 useful Linux commands and utilities guide missed a couple of his favorites:

I’m a NetBSD guy, but I have RHEL at work.

As for commands, I like:

1. pax better than tar

2. hexdump better than od

3. tnfpt better than wget

Pax has a much better command-line interface than tar, especially for copying trees. Consider:

$ pax -rw -pe src dest # to copy a tree

$ pax -wzf file.pax.gz src # to create and archive

hexdump -C is what you almost always want.

Tnftp (a port of the NetBSD FTP client to other systems) is a much saner way to fetch stuff. Why the GNU world focuses on wget instead is a mystery to me. It doesn’t do anything tnftp doesn’t do, and it doesn’t do anything better, either.

If you would like to share your opinions of our essential Linux command guide, feel free to drop us line and share some of your favorite commands with the Enterprise Linux Log.

The AFP reports that the French police force is going all-Linux, completing its switch from Microsoft to Linux by replacing all 70,000 of its desktops with Ubuntu boxes.

That’s a lot of Linux.

Last Friday saw the release of the Linux Kernel version 2.6.24. This new version does not include anything spectacular (Linus’ words, not mine), but does include some updates to network drivers. These improvements will assuredly make some users experiencing network issues happy. You can go to http://www.kernel.org to download and build your new kernel.

But should you? Is there ever a reason to build your own kernel and not just wait for it to come down the pipe from your distribution’s software repositories? For the most part, no, you should not need to build your own kernel for enterprise Linux installations these days. Most kernels that come with stock Linux installations such as Red Hat, SUSE and Ubuntu are fine.

However, some occasions require building a new kernel, for example:

- Licensing changes, which forces one to redact certain bits from the stock kernel (see my previous SVV blog post on 2.6.23 and VMware)

- Support for a new CPU ID that is not yet present in the Kernel

- Performance improvement for servers tasked with performing intense calculations

Got any other good reasons for building a kernel? Send them to us.

While computing in general is in the midst of smassive change with regards to virtualization, 64-bit computing available to Linux and Windows platforms, processing cores and very high RAM quantities, storage is about to go through something big as well, thanks to CleverSafe. 

CleverSafe provides an open-source storage grid that slices data across many datastores through a proprietary algorithm. The figure below shows the basic principle in a greatly distributed sliced-storage grid across 9 international datastores:

While jumping into something like this would require some major planning and a buy-in, enterprises have the option of running just the Slicestors server on one system for the purposes of evaluation. The Slicestors system role would install an iSCSI adapter into your system to connect to storage, and once configuration is complete, normal mount commands would be used for access. Also, Cleversafe works best with CentOS, if you’re interested.

Exploring Cleversafe
Cleversafe is a few years old, but you can expect to see some product maturation soon, as I predict this will become even more popular with enterprises. I will see Cleversafe at Technosium 2008 tomorrow in Santa Clara, California and will be sure to follow-up on this.

We have good news for those awaiting the next version of Ubuntu Linux. The next version for the Ubuntu 8 platform, alpha release 4, is to be relased on January 31, 2008, and the list of bugs is getting smaller every day. Some of the new features for alpha 4 include using Firefox beta 3 as a browser which has some new visual effects and functional features. The alpha 3 website will link to alpha 4 when the version is available. Canonical does not offer support services for the beta releases (but you probably already know that.)

Look for a release candidate in April 2008 for Ubuntu server version 8. 

Anyone who has ever used rpm for package management has at one time or the other had to invoke the command:


rpm -qf PATH_TO_FILE


The above command will tell you what package manages the given file, or more specifically, what package was installed, when it was installed and what put that file on the hard drive. This is an incredibly useful feature when you are debugging a system, attempting to figure out where a file comes from, resolving version conflicts, etc. But many people do not realize that Debian derivatives, such as Ubuntu, also has a similar capability.

In your Debian derivative type:


sudo apt-get install apt-file


This command will install the helpful application apt-file onto your system. Once installed, take a look at its command line options by typing the eponymous command. This is what you will see:

apt-file version 2.0.8.2

(c) 2002 Sebastien J. Grossapt-file [options] action [pattern]

Configuration options:

    --sources-list	-s  	sources.list location

    --cache		-c  	Cache directory

    --architecture	-a  	Use specific architecture

    --cdrom-mount	-d  	Use specific cdrom mountpoint

    --package-only	-l		Only display packages name

    --fixed-string	-F		Do not expand pattern

    --ignore-case	-i		Ignore case distinctions

    --regexp		-x		pattern is a regular expression

    --verbose		-v		run in verbose mode

    --dummy		-y		run in dummy mode (no action)

    --help		-h		Show this help.

    --version		-V		Show version number

Action:

    update			Fetch Contents files from apt-sources.

    search|find		Search files in packages

    list|show		List files in packages

    purge			Remove cache files

As you can see, there are a lot of options. The option you need to mimic the functionality of rpm -qf is the search option. For example, to find out what debian package provides the application gpg simply type:


sudo apt-file search $(which gpg)


When you execute this command you are going to get a lot more results than you probably wanted. That is because the apt-file command does a lazy search and searches for any command that patches the pattern /usr/bin/gpg*. In order to narrow the results, use the –fixed-string option:


sudo apt-file search --fixed-string $(which gpg)


The console should display something similar to the following:

[0]akutz@vault:~$ sudo apt-file search --fixed-string $(which gpg)

gnupg: usr/bin/gpg

gnupg: usr/bin/gpg

gnupg: usr/bin/gpg

I hope this little tutorial on apt-file will make your daily life as a Linux system administrator just a little bit easier!

Canonical just announced the release of Ubuntu 6.06.2 LTS Dapper Drake. This point release keeps Ubuntu’s first LTS release up-to-date with the latest round of patches and bug fixes as we approach their next LTS, Hardy Herron in April of this year (8.04). Features and bug-fixes in 6.06.2 include:

- The ability to use a RAID volume as an LVM physical volume.

- Improved support for NFS mounts on x86_64 (amd64) platforms.

All of the new features and improvements can be found on the Ubuntu mailing list.

I was faced with the decision to implement an additional system on the RHEL 4.x series, or make our first jump to the version 5 releases. I decided to have this additional system to stay on RHEL 4.x because of our support situation. As admins are aware, there are many factors that affect a decision like this one.

RHEL 4.x vs. RHEL 5
RHEL is a stable platform among its competition. At just over a year old, its latest build, RHEL 5, is still new to the scene. But RHEL version5.1 was recently released and has enjoyed initial success thus far. The biggest factor in choosing to remain on the 4.x platform was Red Hat’s recent release of version 4.6, keeping the 4.x a current product. With this release, all of our versions remain within the realm of support, so our internal support requirements have not been impacted by another platform. This also keeps us in line with base configuration of applications that are running on the RHEL systems.

You can’t stay on version 4.x forever!
I know, but we made the decision based on what we can best support internally by not multiplying our scope of platforms. But, the version 5.x test bed is just around the corner, and we will increase our comfort with version 5.x (curiously awaiting a 5.2). At that point we would welcome version 5.x by ceasing the version 4.x installs, and migrating to version 5.x if possible.

What is your strategy?
Do you have multiple versions running in your enterprise? What is your thought process in regards to introducing a new distribution? Share your strategies below in a comment.

The recent acquisition of open source database vendor MySQL by software supergiant Sun Microsystems has many asking if this is a good thing. SearchEnterpriseLinux.com expert Don Rosenberg thinks so. He tells the Enterprise Linux Log why he agrees with Andrew Kutz that this might be the best move for MySQL.

Sun now owns the M in the famous LAMP stack. A good thing? Definitely. Open source fans are always happy to see the success of open source pioneers, such as Monty Widenius and David Axmark who have been with MySQL since 1995. One might have mixed feelings about MySQL being acquired rather than going public, for it would be nice to see some large companies develop in the open source market.

When it comes to open source, Widenius and Axmark played by the rules, initially licensing under the GNU lesser general public license (LGPL) and later under the GNU general public license (GPL). Like Red Hat, they understood the value of a large, evangelical user base that paid no revenue but helped spread the product. But unlike Red Hat, MySQL owns all of its code, adding a proprietary advantage to its strategy. This allows MySQL a proprietary license and companies to embed MySQL in proprietary software without violating open source rules. I suspect this proprietary wrinkle of MySQL was one of the things that made Sun interested.

Sun’s strategy a concern
Sun has always been schizophrenic about open source, as evidenced in their Sun Community Source License (SCSL) and the Sun Industry Standards Source License (SISSL). When it offered a Linux desktop, rather than putting the Linux name on it, Sun stamped it with the Java trademark instead. So it’s not hard to believe that the idea of actually owning MySQL versus merely being an equal user of the source code strongly influenced Sun’s decision.

That being said, I’m a little concerned about how slow Sun has been to warm up to open source licensing. In addition, as Sun was also slow to address the fact that software (e.g. Solaris) was as important to its business as hardware, it took a long time for Sun to wake up to the fact that many of its customers were Linux users.

This large-company lethargy influenced Sun’s open source possessions. It took years for OpenOffice to put up a Web page of add-ons, which were buried in a Sun database that only corporate purchasers would be attracted to. Some years ago Sun hired the leading developers of NetBeans and put them in its Prague laboratory, where they were to extend the functionality and reliability of the NetBeans foundation while Sun added upper layers, some of which were to be enterprise-level and proprietary. But outside developers were faster at add-ons than Sun was, and the young Eclipse (note the irony) from IBM was better at gaining market share and functionality.

MySQL’s future
But I think that MySQL has enough mass and momentum to hold its course, and the slow rate of change in the database industry may be more suited to Sun’s pace. Jonathan Schwartz said on his blog, “MySQL is by far the most popular platform on which modern developers are creating network services,” and Sun did have the long vision some fifteen years ago when it made the McLuhan-esqe proclamation that “the network is the computer.” In this age of cloud computing the network is a better-and-better bet.

The billion-dollar acquisition of MySQL by Sun also illustrates a number of truths about today’s software and open source market. The name of the game is support in one form or another. Oracle is trying to increase its support revenue to match its licensing revenue. Why? Because licensing revenue will eventually drop as open source databases become increasingly common. Sun is already offering paid support for Oracle and Postgres databases; it might as well be the go-to location for support for the most popular of the open source databases, MySQL.

As IT departments discover that database systems are taking an increasing share of the budget, more are discovering open source. Proprietary companies that want to survive will have to do it with better service and lower prices.

MySQL is an example of a disruptive technology. At first, it was too puny for the proprietary databases to notice, and satisfied only small users. But its powers grew, and the size of the companies using it also increased. It cannot match the upper limits of DB2 capability at this point, but it will be interesting to see if and when that day comes. But it doesn’t have to; it is already transforming most of the database market. May Sun invest the money (and employ the open source software developers) to take it to the top.

I had an opportunity this week to install Red Hat Enterprise Linux 4.5 (64-bit version) for a system running a specialized vendor application. I’d like to share my frustration with you so that you can avoid such learning experiences in your future server builds. 

On a Dell PowerEdge 2950 III server, we were unable to assign the Dell Remote Assistance Card virtual media capability or DRAC for a floppy image that contained the array controller driver for the OS install. As we did not have a floppy, our install came to a halt when trying to load the driver:

Installation stopped, now what?
This is not a show stopper. In fact, you have two options that can get this situation resolved. One option is access the files in a floppy image format, extract them to a CD-ROM and make an ISO image with only those files. The other option is hook up a USB floppy drive (the Dell BIOS will make this appear as a normal floppy) with the driver files extracted on a legacy floppy.

Rapid rebuilding process
Should you have multiple systems to re-install either for build or restore process, take the time to determine the quickest way to rebuild a server on your hardware. I’d recommend the CD-ROM ISO image simply because I find it easier to manage files than actual media.