Replacing Windows NT Server with Linux
Copyright (c) 1997 Quinn P. Coldiron
qcoldiro@unlinfo.unl.edu


Table of Contents

Introduction

Section 1--Outlining Our Situation

Section 2-- Why we chose RedHat Linux

• Linux Features
• Reliability
• Cross platform
• Compatibility
• Speed
• Inexpensive
• Low hardware requirements
• Ease of administration
• Simple upgrade path
• Remote control and administration

• Setting up the server
•        Linux installation
•        Samba installation/setup
•        Netatalk installation/setup
• Installing The Cats Pajamas (Getting a DOS network application to run on Linux)
•        Installing Cats on the Server
•        Configuring your Windows clients
•        Configuring the DOS emulator to run Cats
• Installing a RAID
• Daily administration
•        System backups
•        Scheduling events
• Replacing your Desktop OS with Linux

Appendix B--Samba

• Just What is SMB?
• Samba FAQ
• Samba HOW-TO

Network Operating Systems (NOS) have many features and capabilities that allow the Information Systems departments of most organizations to better server your organization and help streamline work flow. Each NOS has different features and a excels in different areas. Traditionally, Novell Netware has been viewed as the best file/printer server, Unix was viewed as the premiere application and database server and recently, Windows NT has come in as a good choice for the smaller networks as a file/printer server or application server. The market has become very aggressive with each of these platforms branching out into each others market area. Microsoft has expanded NT into the midrange server market once dominated by Novell and is trying to get into the high end market which was once filled by Unix venders such as Sun, Hewlett Packard and Silicone Graphics.

I inherited an aging Novell Netware 3.11 server when I began my career at the University of Nebraska Press that was on an under powered Pentium 90 and had older disk drives that were failing. I also wanted to expand into other areas, so I knew this was going to be my first project. When I began looking at my options to replace our aging Novell Netware server, I of course turned first to the just released Windows NT Server 4.0. The marketing materials, magazines and television advertisements all told me that this was the one thing that would solve all my problems. The operating system promised to be easier to setup and maintain than the Netware product it was replacing and it could easily handle the 55 users on my network. Fourteen months later, we are running Linux as our server.

Linux is a completely free reimplementation of the POSIX specification, with SYSV and BSD extensions (which means it looks like Unix, but does not come from the same source code base), which is available in both source code and binary form. Its copyright is owned by Linus Torvalds <torvalds@transmeta.com> and other contributors, and is freely redistributable under the terms of the GNU General Public License (GPL). A copy of the GPL is included with the Linux source; you can also get a copy from ftp://prep.ai.mit.edu/pub/gnu/COPYING.

Linux is not public domain, nor is it `shareware'. It is `free' software, commonly called freeware, and you may give away or sell copies, but you must include the source code or make it available in the same way as any binaries you give or sell. If you distribute any modifications, you are legally bound to distribute the source for those modifications. See the GNU General Public License for details.

Linux is still free as of version 2.0, and will continue to be free. Because of the nature of the GPL to which Linux is subject, it would be illegal for it to be made not free. Note carefully: the `free' part involves access to the source code rather than money; it is perfectly legal to charge money for distributing Linux, so long as you also distribute the source code. This is a generalization; if you want the fine points, read the GPL.

Linux runs on 386/486/Pentium machines with ISA, EISA, PCI and VLB busses. MCA (IBM's proprietary bus) is not well-supported in 2.0.x and earlier versions, but support has been added to the current development tree, 2.1.x. If you are interested, see http://glycerine.itsmm.uni.edu/mca

There is a port to multiple Motorola 680x0 platforms (currently running on some Amigas, Ataris, and VME machines), which now works quite well. It requires a 68020 with an MMU, a 68030, 68040, or a 68060, and also requires an FPU. Networking and X now work. See news:comp.os.linux.m68k

Linux runs well on DEC's Alpha CPU, currently supporting the "Jensen", "NoName", "Cabriolet", "Universal Desktop Box" (better known as the Multia), and many other platforms. For more information, see http://www.azstarnet.com/~axplinux/FAQ.html

Linux runs well on Sun SPARCs; most sun4c and sun4m machines now run Linux, with support for sun4 and sun4u in active development. Red Hat Linux is (as of this writing) the only Linux distribution available for SPARCs; see http://www.redhat.com/support/docs/rhl-sparc/

Linux is being actively ported to the PowerPC architecture, including PowerMac (Nubus and PCI), Motorola, IBM, and Be machines. See http://www.cs.nmt.edu/~linuxppc/ and http://www.linuxppc.org/

Ports to other machines, including MIPS and ARM, are under way and showing various amounts of progress. Don't hold your breath, but if you are interested and able to contribute, you may well find other developers who wish to work with you.

Linux is no longer considered to be in beta testing, as version 1.0 was released on March 14, 1994. There are still bugs in the system, and new bugs will creep up and be fixed as time goes on. Because Linux follows the ``open development model'', all new versions will be released to the public, whether or not they are considered ``production quality''. However, in order to help people tell whether they are getting a stable version or not, the following scheme has been implemented: Versions 1.x.y, where x is an even number, are stable versions, and only bug fixes will be applied as y is incremented. So from version 1.2.2 to 1.2.3, there were only bug fixes, and no new features. Versions 1.x.y, where x is an odd number, are beta-quality releases for developers only, and may be unstable and may crash, and are having new features added to them all the time. From time to time, as the current development kernel stabilizes, it will be frozen as the new ``stable'' kernel, and development will continue on a new development version of the kernel.

The current stable version is 2.0.31 (this will continue to change as new device drivers get added and bugs fixed), and development has also started on the experimental 2.1.x kernels. If 2.0.x is too new for you, you may want to stick with 1.2.13 for the time being. However, the latest releases of 2.0 have proved quite stable. Do note that in order to upgrade from 1.2 to 2.0, you need to upgrade some utilities as well; you may wish to upgrade to the latest version of your Linux distribution in order to obtain those utilities. The Linux kernel source code also contains a file, Documentation/Changes, which explains these changes and more.

Most versions of Linux, beta or not, are quite stable, and you can keep using those if they do what you need and you don't want to be on the bleeding edge. One site had a computer running version 0.97p1 (dating from the summer of 1992) for over 136 days without an error or crash. (It would have been longer if the backhoe operator hadn't mistaken a main power transformer for a dumpster...) Others have posted uptimes in excess of a year. One site still had a computer running Linux 0.99p15s over 600 days at last report.

One thing to be aware of is that Linux is developed using an open and distributed model, instead of a closed and centralized model like much other software. This means that the current development version is always public (with up to a week or two of delay) so that anybody can use it. The result is that whenever a version with new functionality is released, it almost always contains bugs, but it also results in a very rapid development so that the bugs are found and corrected quickly, often in hours, as many people work to fix them.

In contrast, the closed and centralized model means that there is only one person or team working on the project, and they only release software that they think is working well. Often this leads to long intervals between releases, long waiting for bug fixes, and slower development. The latest release of such software to the public is sometimes of higher quality, but the development speed is generally much slower.

As of October 24, 1997, the current stable version of Linux is 2.0.31, and the latest development version is 2.1.59.

I will begin this report by covering the areas that we were concerned about and giving reasons that were chose options that we went with. I will not get very technical in this fist section, but will cover in detail how we implemented each part of the system and give examples of the configuration files in section two.

It was time to begin the search for replacements and I had the same issues to deal with that my predecessors had, plus some new ones. The main issue was compatibility with our order fulfillment and inventory system called The Cat's Pajamas. This system was originally developed on the Data General line of mini and mainframe computers using its interpreted language. The system was ported to PC servers (mainly Novell Netware) when a company called Subject Wills ported the interpreter to the PC platform. It has a character based user interface, stores all its data in flat indexed text files and is pretty fast and stable on the Novell platform. Our new requirements were for remote access for our warehouse, and if possible, access for the five Macs that we have, but spend as little money as possible.

I was not very interested in upgrading our current Netware license, since at this time Novell was playing the CEO of the Month game which had caused me to lose much of my confidence in the company. The next logical choice was Microsoft Windows NT. I called Cats and asked if the application would run on NT and they told me that they had a few installations and things seemed to be running fine, but I would need a different version of the DBC interpreter, which they uploaded for me. I proceeded to copy the entire system and database to the NT server and logged on from my Windows 95 PC. I had to rewrite the batch file that Cats uses to start because they were using Netware network commands, but I was able to get them changed and get a session up.

At this point I thought this looked very good, so I ordered the production server. The server was a dual CPU Pentium 150 with built in Adaptec 2940UW SCSI controller and 256MB of RAM. I thought this would be perfect to act as the CATS server, file server and printer server for 50 people, especially since Netware was doing this on a Pentium 90 with 64 MB of RAM. I got the server a few weeks later, installed NT 4.0 and Cats. After I got Cats installed I had our business department and Customer Service department try running some large and CPU intensive reports on it and things seemed to run great, so we pushed it into service. At this same time I was setting up a RedHat Linux 4.1 machine on a Pentium 100 with 32 MB of RAM. I had used RedHat Linux in my previous position with great luck as a web server and StreamWorks audio/video server so I wanted to see if it could fit in here, but I didn't know where or how.

Life after moving Cats to NT was a nightmare. The system was crashing two to three times a day with no reason that I could find. I was on the phone with Microsoft and Cats constantly, but nobody could figure it out. Microsoft had me apply Service Packs one through three and a few HotFixes, which helped, but it still was crashing at least twice a week with the infamous "Blue Screen of Death". After many weeks and about $1500.00 in phone support from Microsoft, the technical support rep told me that I should find a better software package than The Cat's Pajamas. This is was not the solution I was looking for, since this is the package that a sizeable percentage of presses our size nationwide are running, so I was forced to bring the old Novell server back into production until I could figure something out.

By this time I had upgraded the RedHat Linux machine to version 4.2, but it still wasn't doing much, so I added two IDE 1.6 GB hard drives, installed SAMBA and copied CATS over to this computer to play a little. I was able to connect from my Windows 95 computer and run Cats without any modifications to the batch file I had created for the NT server. I had the Business manager and Customer Service manager try the reports again and everything seemed fine, but I still had my doubts, so I just kept the system up for myself and the IS department to test until a later date.

The Netware server was still holding on and doing its job until I got a call one Thursday night at about 9:00 from our Customer Service manager. He had been running the reindexing program to prepare for the month-end closing that was to start Friday, but the server had crashed. I worked on it until about 12:30 am and finally got it working, but the whole thing died again Friday morning at about 6:30 and again at 7:00, so I knew we were in trouble. The only option left was to replace the server now and the only thing I had was the Linux server, so I restored Cats off our Windows NT tape backup machine to the Linux server and changed the login scripts to connect the users to this server for the Cats drives. Within one hour, we were back in operation.

After completing the morning duties, we normally run a complete Cats backup before we continue with closing which usually would take two hours to complete on the Netware server. The Linux machine was able to do the entire backup in 45 minutes, cutting a little over an hour off our closing time. This increase in speed came from a decrease in hardware because the Linux server was running only 32 MB in RAM and IDE hard drives where the Netware server had 64 MB in RAM and SCSI drives. The speed increase has been noticed in daily work also. I get almost daily remarks that the system seems to be running faster and more reliable.

We have recently upgraded the CPU to a 200Mhz Pentium and have upgraded the Memory to 64 MB to handle the newest plans of making this server replace our Windows NT file/printer server, which still crashes about twice a month for no reason, even after an additional $1,500 in tech support with Microsoft. This single computer running RedHat Linux will replace both our Novell Netware 3.11 server and our Windows NT 4.0 server, while decreasing total hardware requirements. With the recent advances from the Samba team in supporting the NT domain structure and the December 1997 release of RedHat 5.0, I expect to have a very efficient and inexpensive server for our Windows 95, Windows NT and Macintosh clients.

• · multitasking: several programs running at once.
• · multi-user: several users on the same machine at once (and no two-user licenses!)..
• · multiplatform: runs on many different CPUs, not just Intel.
• · multiprocessor: SMP support is available on the Intel and SPARC platforms (with work currently in progress on other platforms), and Linux is used in several loosely-coupled MP applications, including Beowulf systems (see http://cesdis.gsfc.nasa.gov/linux-web/beowulf/beowulf.html) and the Fujitsu AP1000+ SPARC-based supercomputer.
• · runs in protected mode on the 386.
• · has memory protection between processes, so that one program can't bring the whole system down.
• · demand loads executables: Linux only reads from disk those parts of a program that are actually used.
• · shared copy-on-write pages among executables. This means that multiple process can use the same memory to run in. When one tries to write to that memory, that page (4KB piece of memory) is copied somewhere else. Copy-on-write has two benefits: increasing speed and decreasing memory use.
• · virtual memory using paging (not swapping whole processes) to disk: to a separate partition or a file in the filesystem, or both, with the possibility of adding more swapping areas during runtime (yes, they're still called swapping areas). A total of 16 of these 128 MB swapping areas can be used at once, for a theoretical total of 2 GB of usable swap space. It is simple to increase this if necessary, by changing a few lines of source code.
• · a unified memory pool for user programs and disk cache, so that all free memory can be used for caching, and the cache can be reduced when running large programs.
• · dynamically linked shared libraries (DLL's), and static libraries too, of course.
• · does core dumps for post-mortem analysis, allowing the use of a debugger on a program not only while it is running but also after it has crashed.
• · mostly compatible with POSIX, System V, and BSD at the source level.
• · through an iBCS2-compliant emulation module, mostly compatible with SCO, SVR3, and SVR4 at the binary level.
• · all source code is available, including the whole kernel and all drivers, the development tools and all user programs; also, all of it is freely distributable. Plenty of commercial programs are being provided for Linux without source, but everything that has been free, including the entire base operating system, is still free.
• · POSIX job control.
• · pseudoterminals (pty's).
• · 387-emulation in the kernel so that programs don't need to do their own math emulation. Every computer running Linux appears to have a math coprocessor. Of course, if your computer already contains an FPU, it will be used instead of the emulation, and you can even compile your own kernel with math emulation removed, for a small memory gain.
• · support for many national or customized keyboards, and it is fairly easy to add new ones dynamically.
• · multiple virtual consoles: several independent login sessions through the console, you switch by pressing a hot-key combination (not dependent on video hardware). These are dynamically a located; you can use up to 64.
• · Supports several common filesystems, including minix, Xenix, and all the common system V file systems, and has an advanced filesystem of its own, which offers filesystems of up to 4 TB, and names up to 255 characters long.
• · transparent access to MS-DOS partitions (or OS/2 FAT partitions) via a special filesystem: you don't need any special commands to use the MS-DOS partition, it looks just like a normal Unix filesystem (except for funny restrictions on filenames, permissions, and so on). MS-DOS 6 co pressed partitions do not work at this time without a patch (dmsdosfs). VFAT (WNT, Windows 95) support is available in Linux 2.0 ·special filesystem called UMSDOS which allows Linux to be installed on a DOS filesystem.
• · read-only HPFS-2 support for OS/2 2.1
• · HFS (Macintosh) file system support is available separately as a module.
• · CD-ROM filesystem which reads all standard formats of CD-ROMs.
• · TCP/IP networking, including ftp, telnet, NFS, etc.
• · AppleTalk server
• · Netware client and server
• · Lan Manager (SMB) client and server
• · Many networking protocols: the base protocols available in the latest development kernels include TCP, IPv4, IPv6, AX.25, X.25, IPX, DDP (AppleTalk), NetBEUI, Netrom, and others. Stable ne work protocols included in the stable kernels currently include TCP, IPv4, IPX, DDP, and AX.25.

My Linux server was initially setup in January 1997 as a test platform to see if it could handle being a production server. At this time I had installed a copy of The Cat's Pajamas, Apache, StreamWorks (audio/video server) and Samba. Using Samba, I was able to connect the Information Systems Department to this server so we could run Cats and test the audio/video server. A typical day of testing the system would include reindexing and reformatting Cats while serving multiple 100 megabyte video files and acting as a file server. The Linux machine was able to handle all of this while running on only 32 megabytes of RAM and using a Pentium 100 CPU.

From January 1997 to July 1997, we experienced only three shutdowns of this server, two of which were caused by power failures in the building and the third time was due to stupidity on the administrator's part. This reliability was a key factor in our judgment to use Linux as our server platform.

The installation of NT was simple, and it was easy to get the system connected to all the printers, but it proved to be unreliable. We were experiencing crashes every day, even after installing service packs one and two. I then installed a hot fix to service pack two which was to fix errors in the Macintosh services which helped. I was down to crashing just one to two times a month. I also tried to run Cats off this system, which was a complete disaster since we could not get more than five users at a time in Cats, and reports would take forever to run.

Some of Samba's server cousins include DEC Pathworks, Microsoft LAN Manager/X, OS/2 Lan Manager, IBM LAN Server, Syntax Server and Windows NT Server. Some clients such as Windows 95/Workgroups or Warp Connect can also act as low volume servers with limited management facilities.

SMB is becoming very popular, mainly owing to these factors:

• · Windows 95 has dial-up access to PPP servers with an included service, and this service allows one to "browse" to public shares on the Internet.
• · Samba is "free" and this is a lot less expensive than Novell! (friendlier too!)
• · With Samba, Unix servers, well connected to a global network, can speak in a "native" protocol of clients. It is much simpler to maintain one more protocol on a capable server than teach new tricks to multiple clients which were never meant to do something else.
• · There is an established, well tested way of doing SMB over TCP/IP described in publicly available RFC 1001 and RFC 1002 documents. This means that SMB has a head start when it comes to Internet integration.

On the Macintosh side, Netatalk provides the connectivity services. Netatalk is a Unix implementation of the AppleTalk Protocol Suite, originally for BSD-derived systems. It includes support for routing AppleTalk, serving Unix and AFS file systems over AFP (AppleShare), serving Unix printers and accessing AppleTalk printers over PAP. A number of other minor printing and debugging utilities are also included.

My hopes were verified when we formally flipped the switch on the Linux server and turned off the Novell server. Cats ran smoothly and gracefully scaled up to the normal daily user level without any slowdowns or misfortunes. The Novell Netware server would normally experience periods when the system would slow down severely, and then speed back up to normal levels after 5 to 10 minutes. During these slowdowns, Cat's would become virtually useless, and sometimes users would be booted out of Cat's totally.

Another key area that I needed to be compatible with was the world Wide Web. I needed to be able to get CGI programs that were already tested and in use without having to develop them all myself. I also wanted to be able to get help easily and cheaply on setting up the web server and fixing it when it breaks. Linux uses Apache, the most widely used web server in the world. I can easily get support from mailing lists , news groups and Internet BBS's.
(http://www.netcraft.com/survey/)

Filesystem 1024-blocks Used Available Capacity Mounted on

/dev/hdc1 2417493 1584580 707923 69% /usr/local/samba-sys

Technical support is no the only area of expense involved with Windows NT. The table below shows typical applications and the cost for each platform.
 

Our Windows NT 4.0 server:

Unix usually gets the standard rap of being difficult to manage and requiring a steep learning curve. While many of the tools that will be used to manage the system are still command line based, a growing number are graphical and comparable to Windows in ease of use. This is especially true with the RedHat distribution, which includes a graphical control panel which allows the administrator to change system setting without working through the many text files which store this information. The control panels I use the most are the ones which setup the printers, file system, users and packages. If you use the included RedHat Package manager to install new applications, they are listed in the package manager, which allows you to graphically see what applications are installed on your system and remove them if wanted, much like the uninstall option on Microsoft Windows.

While it is not required that you uninstall the previous version of Windows NT before installing the current release (4.0), it is recommended. This can be a very intensive installation, since you would need to setup all the user accounts again, along with reinstalling any software and automated systems and scripts.

When I upgraded from RedHat 4.1 to 4.2, the installation when smoothly whiteout any problems. While this was not a major revision step, RedHat has assured me that the soon to be released version 5.0 will offer a smooth transition from 4.2.

From the RedHat Linux manual:
 

1.5 Upgrading from a Prior Version  The installation process for Red Hat Linux 4.2 includes the ability to upgrade from prior versions of Red Hat Linux (2.0, 2.1, 3.0.3, 4.0, and 4.1) which are based on RPM technology. Upgrading your system installs the modular 2.0.x kernel as well as updated versions of the packages which are installed on your machine. The upgrade process preserves existing configuration files using a .rpmsave extension (e.g., sendmail.cf.rpmsave) and leaves a log telling what actions it took in /tmp/upgradelog. As software evolves, configuration file formats can change, so you should carefully compare your original configuration files to the new files before integrating your changes.  To upgrade a Red Hat Linux system, you must use a boot floppy (and possibly a supplemental floppy), just as if you were performing a full install. After selecting the medium to install from (and setting up networking information if it's required), the installation procedure prompts you to either Install or Upgrade; select Upgrade. This upgrade procedure is the functional equivalent of running the upgrade script in prior versions of Red Hat Linux.  Please note that some upgraded packages may ``depend'' on other packages which may not be installed on your system. The upgrade procedure takes care of these dependencies, but it may need to install additional packages in order to satisfy them.

Telnet is text only, so this will be a drawback for many, but luckily for us, Linux supports X Windows. X Windows can be used locally on the server's console, like Windows NT, but can also be run remotely from any computer that can run an X client. These remote terminals can by other computers running Linux, or any other flavor of Unix and many X clients can be downloaded or purchased for Windows, Windows NT and Macintosh computers. Running X terminals does require more bandwidth, so I prefer just telneting into the server.

Another nice feature is the ability to display results of various system utilities in a web page. One of the things our Customer Service manager likes to know is which users have which files open and locked while using Cats. I wrote a simple Perl script to run the smbstatus program and print the results as a web page.

The output web page:
 

Windows NT provides a graphical program to show a chart us system performance, but it is recommended that you do not run it on the server that you want to monitor, because the monitor program itself takes too much of the system resources, and thus skews the data. Linux provides a system monitoring tool called TOP that is character based so it can be run on the server. Since it is character based, you can telnet into the server and run it from anywhere, getting an accurate picture of how the server is running.

Sample output of TOP:
 

• · hard drive(s): number and size; if you have more than one, it's helpful to know which one is first, second, etc.
• · memory: amount of RAM.
• · CD-ROM: its interface type (IDE, SCSI, or other interface), and, for non-IDE, non-SCSI CD-ROMs, the make and model number.
• · SCSI adapter: make and model number.
• · network card: make and model number.
• · mouse: type (serial, PS/2, or bus mouse), protocol (Logitech, Microsoft, MouseMan, etc.), and number of buttons; also, for serial mice, the com port it is connected to.
• · If you will be connected to a network, be sure you know your IP address, netmask, gateway IP address, name server IP addresses, domain name, and hostname. If you don't know these values, ask your network administrator.
• Also, if you will be installing the X Window System, you should be familiar with the following:
• · your video card: make and model number or video chipset, amount of video RAM.
• · your monitor: make and model number, allowable range of horizontal and vertical refresh rates.

Installing from CD-ROM or via NFS requires only a boot diskette. Installing from a hard drive, via FTP, from an SMB volume, or from a PCMCIA device (including PCMCIA-based CD-ROMs) requires both a boot diskette and a supplemental diskette. Section 2.3.1 below explains how to create boot and supplemental diskettes.

CD-ROM

If you have a Red Hat Linux CD and a boot diskette you will need a supported CD-ROM drive and either a 3.5 inch floppy drive or an operational installation of MS-DOS on your machine in order to run the installation utility. If a boot diskette did not accompany your CD, you will need access to a computer running either Linux or MS-DOS to create a boot diskette from the CD.

NFS

If you wish to install over a network, you will need to mount the Red Hat Linux CD-ROM on a machine that supports ISO-9660 file systems with Rock Ridge extensions. The machine must also support NFS. Export the CD-ROM file system via NFS. You will need to have name services configured, or know the NFS server's IP address, and the path to the exported CD-ROM.

FTP

For an FTP install, you must have a boot disk and supplemental disk. You will need to have a valid name server configured or the IP address of the FTP server you will be using. You will also need the path to the root of the Red Hat Linux directory on the FTP site.

SMB Shared Volume

If you wish to install from an SMB shared volume, you will need to mount the Red Hat Linux CD-ROM on a Microsoft Windows NT or Windows 95 server that supports shared volumes. You will need to have name services configured, or know the server's IP address; you will also need the name of the shared volume containing the Red Hat Linux CD-ROM and the account number and password to use to access the volume.

Hard Drive

To install Red Hat Linux from a hard drive you will need the same boot and supplemental disks used by the FTP install. You must first create a RedHat directory at the top level of your directory tree. Everything you will install should be placed in that directory. First copy the base subdirectory, then copy the packages you want to install to another subdirectory called RPMS. You can use available space on an existing DOS partition or a Linux partition that is not required in the install procedure (for example, a partition that would be used for data storage on the installed system).

If you are using a DOS filesystem, you may not be able to use the full Linux filenames for the RPM packages. The installation process does not care what the filenames look like, but it is a good idea that you keep track of them so you will know what you are installing.

PCMCIA

If your CD-ROM, Ethernet card, or local hard disk is connected to a PCMCIA adapter, you must install with PCMCIA support. You need a supported PCMCIA controller and a supported PCMCIA SCSI adapter or Ethernet card. Installing via PCMCIA requires the use of a supplemental diskette.

Please See Appendix A for The Official RedHat Installation Guide.

An advantage to installing Samba using RPM is that it is easier. A disadvantage is that the versions are not as current, but they are usually more stable and reliable. I choose to go directly to the Samba web pages and download the most current non-beta release.

The picture on the left is the Network Neighborhood window on my NT workstation and is currently showing everything that I have access to on my Linux server.

If you installed Samba using RPM, it will automatically be started on system startup, but I had to manually add it to my rc.local script to have it start on bootup because I did not install from the RPM. I chose to install from the tarballs available at the official Samba site, so I could always have a current installation. To start the server manually, I just issue these commands:

/usr/local/samba/bin/smbd -D

/usr/local/samba/bin/nmbd -D

Extensive has been accomplished that allows Samba to act more like an NT server, including making a Linux server a Primary Domain Controller and more features are currently being added. I have listed the steps that I took in making my Linux server the Primary Domain Controller (PDC) for our network.

1. Set up the server.
       a. Create the smb.conf file.
       b. Create the netlogon directory. I put mine in /home.
       c. Restart Samba.

2. Setup the Windows 95 clients. (We set our up for roaming profiles.) Do not reboot until step 3!
       a. Control Panel
               1. Click Passwords, then Profiles and choose these settings:
                       a. Users can customize their settings.
                       b. Include desktop items.
                       c. Include start menu.
               2. In the Control Panel, select Network.
                       a. Under Identification, set your WORKGROUP name.
                       b. Access control = USER-LEVEL.
                       c. Obtain the list from YOUR SERVER NAME.
                       d. Under Client Configuration
                               1. Select CLIENT FOR MICROSOFT NETWORKS.
                               2. Under that client's properties.
                                       a. Log onto Windows NT domain.
                                       b. Enter your WORKGROUP name.
                                       c. Select LOG ON AND RESTORE CONNECTIONS.
                       e. Set primary network logon to CLIENT FOR MICROSOFT NETWORKS.
       b. Set up the user profiles.
               1. Install the Policy Editor from the Windows 95 CD.
               2. Create a new profile called config.pol and save it into the NETLOGON share on the Linux server. Make sure you add all the users on your server! This step only has to be
                   done once, not on each client.
               3. Using the Policy Editor, select Open registry from the File menu and select the options you want.
               4. Save it as config.pol and copy into the netlogon share on the Linux server. Set its permissions as 755.

3. Reboot the Windows 95 computer(s) and logon.

4. Possible problems.
       1. Windows95 OSR2 seems to send the passwords as uppercase, so try changing the passwords on the Linux server to uppercase.
       2. I had problems making my Linux server the WINS server, so you have some choices to fix this:
               a. If you do have another NT server, make it the WINS server and point all the clients to it.
               b. Disable WINS on each client and edit the c:\windows\Lmhosts.sam file and add the server and workstations to this file. Save the file as Lmhosts, NOT Lmhosts.sam This file can be appended to from a server, but
                   I haven't tried doing it from Linux/Samba. The entries look like this: xxx.xxx.xxx.xxx computername
               c. Disable WINS. At this time, we are running our Win95 clients with WINS disabled, but we do have a DNS server.

My Samba configuration (smb.conf) file looks like this:
 

The specific portion of the smb.conf file for Cats is listed here:
 

I have written a CGI Perl script to make it simple for our Business Department and Customer service department to lock Cats during month-end closing and reindexing procedures. The script asks for a password, and then replaces the standard batch files to start Cats with another batch file that says Cats is locked for closing or maintenance. When they are finished, they can unlock Cats by running the unlock CGI script. The scripts are listed here:

Lock Cats:

DOSEMU uses an image file of a hard disk to emulate a DOS hard drive so you do not need to have a DOS partition. Using the /var/lib/dosemu/setup-hdimage program will step you through making a basic hdimage file. The only required item is a bootable DOS 6.22 floppy disk with whatever DOS tools you think you might need. I included EDIT.COM and QBASIC.EXE. After running through the setup-hdimage program, start the emulator with xdos or dos, depending on if you are in X-Windows or not. The default setup gives you the Linux hard drive as DOS drive D:\. To copy the programs I wanted off the floppy to the hdimage file, I opened another xterm window and mounted the floppy to /mnt/floppy, then switched back to DOSEMU, changed to D:\mnt\floppy and copied EDIT.COM and QBASIC.EXE to C:\. I could now run the MS-DOS editor to edit the config.sys and autoexec.bat files. If I wanted, I could also write a quick QBASIC program.

DOSEMU gives you a utility called LREDIR which redirects Linux directories to DOS drive letters. Below is the AUTOEXEC.BAT that I use with DOSEMU to give me the drive letters for Cats and for Robotronics.
 

The SCSI card is a PCI card with a caching module added to it. The cache module has four SIMM slots on it and will accept any standard 32 pin SIMM up to 16 meg each for a total of 64 megabytes of cache. The card comes with a four meg module already installed, and that is what we currently are using.

The storage subsystem requires you to supply your own drives and comes in two style. One supports narrow SCSI drives and the other supports wide drives. We are using the narrow drive cabinet, mainly because I already had three drives that I wanted to use. Installing the drives can be tricky, as the data cable is not color coded and I could not tell what wire was for pin one. The first time around I had the drives installed backwards, but nothing was hurt. I called the tech support line and they were very helpful in getting the problem fixed.

After installing the card in the server, the next step is to configure the RAID with the RAD level of choice. We chose to run RAID level 5. While Linux does have built in drivers for the DPT controller card, DPT does not offer Linux utilities to configure the unit. What I did was temporarily plug a 200 meg IDE hard drive into the server to boot from and ran the DOS configuration utilities which are graphical and easy to follow. Using the mouse, I selected the three drives that I wanted to include in my RAID, then selected the RAID level I wanted, saved the configuration and shutdown the computer. I then unplugged the IDE DOS drive and reconnected the IDE Linux boot drive.

I kind of cheated when I installed the driver for the SCSI card by rerunning the Linux setup program. I had other things that needed fixing, so I thought this would be the best way. I could have just added the daemon by using the tools in the control panel as well or by typing insmod eata-dma at the command prompt and the driver would be loaded and I could see the report that the system now recognizes my SCSI card and RAID. During the installation, select YES when asked if you have a SCSI card and select correct driver (EATA-DMA) for the card you have. When you are finished and reboot the computer, you will see the RAID initialize and report the settings you configured under DOS, then the kernel recognizing the card.

Now, you need to create a partition and format the drive. To make a partition, use the fdisk command and follow the menu to make a Linux primary partition. Formatting the RAID is as easy as formatting any other drive. Remember that the system sees your RAID as a single, large disk. To format it in Linux, issue the mkefs2 /dev/sda1, or whatever SCSI drive it is. To actually use the disk after formatting, you need to mount it someplace. I wanted to use this disk for the home directories of our users, so I mounted it as /home/raid. But, whenever I added a user, they were still being added to /home and not on the RAID, so I edited the /usr/sbin/adduser script (which is just a Perl script) and changed the home directory location to /home/raid. Now, whenever I add a user, they are added to the RAID.

another SureStore 6000 tape drive and Cheyenne Arcserve software. I have also been known to make a quick backup to the RAID using tar. Tar is a simple utility that has been around Unix almost as long as Unix has been around.

To make a tar backup, just issue the following command:

tar cvf archivename.tar /directory-to-archive

This will create a tar file named archive.tar of the contents of the directory directory-to-archive. While this works fine, the resulting file might be very large because we did not apply any compression. By

modifying the command to look like this:

tar cvzf archivename.tar.gz /directory-to-archive

We compress the tar file with gzip after the file is made. At a later date, we might want to view what is in the tar file, so we can issue the tar tvf archivename.tar.gz to list the files in the archive.

While backing up to disk is OK for a temporary thing, you should always backup to tape for critical data. Linux supports two types of tape drives. The first is the style that connects to the floppy drive

controller, like the Colorado drives and Iomega Ditto. The other style connect to a SCSI controller. Floppy tape drives have the device names of /dev/ft0, dev/ft1 and so on. SCSI drives are /dev/st0, /dev/st1 and so on. These devices are "rewinding" devices which rewind the tape when the operation is completed. IF you are archiving more than one session to the tape, then use the non-rewinding devices /dev/nft0, /dev/nrft1, /dev/nst0, /dev/nst1 and so on.

After writing the archive to tape, you can use the mt (magnetic tape) command to rewind, retention and find sessions on tape. The table below shows the commands.

mt /dev/nft0 rewind

mt /dev/nft0 retention

mt /dev/nft0 fsf 1 skips the current session to find the next session on tape.

To use the mt command, you must use the non-rewinding devices.

Using tar to make backups has its good points and its bad points. On the bad side, neither tar or gzip are fault-tolerant. Compressing tar files with gzip will greatly reduce the amount of space on the backup

media, but if even one block of the archive gets corrupted, which can happen on tapes, the entire file can be useless. Usually, you will be able to retrieve all data up to that point, though. A better solution is to use a "real" backup system such as BRU (Backup and Restore Utility) that comes with a purchased version of RedHat Linux 5.0. Backup systems usually compress each file individually, so that if the media is damaged, you will not usually lose the entire archive.

BRU has both a command prompt interface and a graphical X-Windows interface and is included with RedHat 5.0. BRU offers automated backup scheduling and scripting, so it is very easy to setup a simple but effective and reliable backup system. I find the system as easy to maintain as ArcServe on Windows NT.

You might notice the entry

# Check to see if network volumes are mounted (at 10:00 p.m.).
# These need to be mounted since this machine performs the
# backup at 11:55.
0 22 * * 1,2,3,4,5 checkmounts

This calls a script I wrote to check that the NT server is mounted so I can backup all data on it. I mount the server using the smbmount command which allows a Linux machine to mount shares from a Windows machine. This script just verifies that a certain file or directory is actually at the mount point. I use these ``reference'' files for checking purposes, so they are read-only, so they can't be accidentally deleted by users. The script is listed here:
 

• RedHat
• K Desktop
• Amazing Linux Desktops

I have found that, for the most part, I can replace Microsoft Office with Applixware. Applixware is an office suite for Linux (an many other Unix platforms) that can read and write Word and Excel files, which allows me to share documents with the rest of the press. While I still cannot read Access databases, I am working out a solution for that which we will discuss later.

Applix Words

• · Live links to Applixware or third-party applications
• · Mail-enabled compound document technology
• · Landscape, portrait and merge printing
• · Multiple undo and redo
• · Hypertext with optional HTML output
• · Forms editor
• · International dictionaries, thesauri and spell checkers
• · Conditional text and cross-referencing; find and replace
• · Embedded equations and calculation support
• · Multiple typefaces, type styles and sizes
• · Highly graphical user interface
• · Tables and frames; borders and shading
• · Multimedia support including audio and video

• · Live links to other Applixware and external applications
• · Calculation-based attributes
• · 3D spreadsheets and charts
• · Integration of external functions
• · Built-in string, financial, mathematical, and statistical functions
• · Goal seeking
• · Multiple named views
• · Dynamic inter-spreadsheet links
• · Live graphs and charts
• · Editable graph styles
• · Print preview
• · Borders and shading
• · Drag-and-drop moving of cells
• · Projection tables
• · Relative and absolute cell references
• · Built-in database functions
• · Built-in shared library, RPC, and shell command mechanisms
• · True minimal recalculation

• · Feature-rich palette, fill patterns & drawing tools
• · Horizontal and vertical shearing
• · User-definable fill patterns, brush styles and shapes
• · Unlimited color support and point sizes
• · Rotating and scaling options
• · Drop shadows
• · Text wrapping
• · Image digitizing
• · Pixel editing
• · Managed layers and animation

• · Fully mail-enables all Applixware products
• · Personal and global user lists, distribution lists and aliases
• · Sorting, filing and querying on incoming messages
• · Automatic conversion of messages/documents to recipient's preferences
• · UNIX sendmail transport and MIME compliant
• · Shared folders

• · Availability of a wide range of basic and advanced HTML concepts
• · More than 25 standard HTML styles for headlines, subheads, lists and text
• · Auto-conversion of Applix Words styles to standard HTML styles
• · Auto-conversion of Applix Graphics and Applix Presents to GIF, including linked or embedded graphics
• · Auto-conversion or creation of tables
• · Full control over background color, text and hyperlink colors
• · Auto-bulleting and auto-listing functions
• · Dialog boxes and pull-down menus for all attributes and activities
• · Straightforward linking and editing of existing GIF files

• · Robust interpretive programming language
• · TCP/IP socket interface and remote procedure calls (RPCs)
• · Interactive debugger, compiler/interpreter
• · Arithmetic, string and Boolean operators
• · Over 2500 built-in macros. Time-saving features for all users
• · User-definable menu bars, dialog boxes and keybindings, menu-bar editor across all Applixware products
• · Keystroke and mousepick recorder
• · Live linking and launching of external applications from Applix Words,
• · Spreadsheets, Presents, and Mail.
• · User-definable ExpressLine icons for automating tasks
• · Incremental zoom

This new version features advanced HTML authoring tools for adding pages to your website simply and directly. It also has dramatically improved filter technology enabling you, for example, to import and work on Word or WordPerfect files and export back in those formats. If you want the ``real thing'', Corel has version 7.0 of Word Perfect available for Linux.

The graphical user interface for Linux (and Unix in general) is X Windows. A typical X Windows setup is shown below. This system is running the Common Desktop Environment (CDE) window manager. X Windows is divided into two subsystems consisting of a server and a client and you are free to change either of these as your want. The main reason for changing an X server is to gain better speed and better support for your video cards and monitors. The main reason for changing window managers is to get the look and feel you desire. You can not change the server or window manager under any of the Windows products, because Microsoft has decided the look and feel you get. The closest they get is the Plus! pack and its themes.

Another possibility , which will save money and breath life into your old 486's is to install Linux on them and use them as a NC (Network Computer). Linux's GUI is X, so it is perfect for an X terminal (with either another Linux machine as the server, or any other Unix system) and you can run Insignia Solutions Inc.'s Keoke client under Java in order to turn your Linux workstation into a thin client that runs Windows (virtual NT) applications.

As a rule, Linux performs radically better than Windows 3.1 given the same amount of RAM. Linux can run complete with a graphical desktop on a 386 with 4MB of RAM, if you're patient. Add Netscape Navigator, and you can get away with a 486 and 8MB of RAM. Consider 16MB of RAM a performance sweet spot, especially if you use Communicator. Linux also has unimposing disk space requirements. The bare-bones Caldera graphical client uses about 68MB of disk space. You'll want to install a few more packages than that (Netscape, Java, and perhaps a friendlier window manager than the default Fvwm), and you'll need at least a 32MB swap partition. When you consider all the factors, a 200MB disk is plenty for most desktops.