Getting the Right Start in Configuring Apache

The primary difference between the Apache-based server and the original HTTP server is the control you have over the server's features. This control is implemented by more options and parameters for each server directive. It is extremely important to understand what each directive is doing and what options are available for you to use. Train yourself to configure this server using a text editor (e.g., Notepad or WordPad) and learn the directives before using the GUIs. To implement many of the advanced functions available within the Apache-based server, you need to thoroughly understand what each directive does. The Apache Software Foundation's Web site (www.apache.org) is the best source for this documentation. In addition, IBM has published two online PDF documents containing useful Apache-based information:

• The "IBM HTTP Server for AS/400" online reference manual
• The "HTTP Server (powered by Apache) Directives Organized by Module" online reference manual

What Is a Virtual Server, and Why Do I Need One?

For each server instance, OS/400 Apache runs a series of jobs (usually four or more jobs) in the QHTTPSVR subsystem. You can view these jobs by performing a Work with Subsystem Jobs (WRKSBSJOB) command, as follows:

WRKSBSJOB SBS(QHTTPSVR)

To find your server jobs in this subsystem, look for the name of your server instance as the OS/400 job name. A server daemon receives all TCP/IP requests from users. The daemon locates a thread or causes a thread to be dispatched in one of the helper (or processing) jobs and passes the request to the job and thread for processing. Each server instance consumes a significant amount of resources and introduces overhead on your system.

If you need to support multiple Web sites in OS/400 Apache, you may create a server instance with its own configuration file for each Web site uniform resource identifier (URI). (Note: URI is synonymous with URL but more accurate, encompassing a broader scope than the HTTP-associated with URLs.) This approach works well, but it requires that each server instance be configured to use a unique port number or a unique IP address. Separate server instances can also add overhead to your system as I just demonstrated. That's where virtual servers and virtual hosts come in.

Before You Virtual Host

Before exploring virtual servers, it is important to understand how the server connects to IP addresses and ports. All TCP/IP servers connect to the communications subsystem by creating a programming device called a socket. The socket is the interface between an application program (our OS/400 Apache-based server) ) and the telecommunications subsystem (TCP/IP). When a server starts, it issues a BIND command with various options that control which IP address(es) and port(s) the server will listen to for messages from the network.

When you configure the Apache-based server with the Listen 80 directive, you are telling the server to create a socket that will receive messages addressed to port 80 for all of the IP addresses defined to your system. An iSeries machine running OS/400 V5R1 can support up to eight communications adapters with 2000 IP addresses defined for each adapter, up to a maximum of 16,000 IP addresses. The Listen 80 directive will route any traffic on any of the 16,000 addresses to port 80. This may be desirable sometimes, but it is not what most users want. If you want to configure your server to accept messages that only address port 80 on the 10.130.39.223 IP address, you have two options. You can enter a Listen directive for 10.130.39.223 that explicitly specifies the IP address and port number to listen for, as follows:

Listen 10.130.39.223:80

Or you can enter a generic Listen directive for the 10.130.39.223 address followed by a Port directive that specifies the port numbers to listen to. This configuration would look like the following example:

Listen 10.130.39.223
Port 80

It's important to note that the Port directive in the second configuration is only valid when there are no port numbers specified in your Listen directive. If you use the port number parameter in a Listen directive, the Port directive for any IP addresses that are specified in the Listen directive is ignored.

The IBM HTTP Server (powered by Apache) offers extreme flexibility in choosing network connection options. Unlike the original HTTP Server for AS/400, the Apache-based server allows you to specify lists of IP addresses that you want your server to listen for. To specify that the server should listen to ports 80 and 8081 on three separate IP addresses, enter this group of directives in your server instance:

Listen 10.130.39.223
Listen 10.130.39.225
Listen 10.130.39.227
Port 80,8081

Using these directives, each of these IP addresses would act as a virtual server for your Apache-based server instance, and your server would answer any HTTP requests on your network for files at these particular IP addresses and port numbers. If you code these directives in any of the sample configurations that were listed in my earlier article "OS/400 Apache Has Arrived," then a request from a browser targeting any of the supported IP addresses will receive exactly the same pages from the same directory structure.

Why Use Virtual Hosting?

Virtual hosting is used to share a single server instance (one daemon job and one set of request processing jobs) among several URIs. It provides a technique for allowing the instance to support multiple, individual Web sites, each with its own domain name, IP address, and content. This approach reduces the number of configuration files you must maintain, and it reduces OS/400 system overhead because each site shares the same OS/400 processing jobs. The HTTP Server (powered by Apache) supports four types of virtual hosting:

• IP-based virtual hosting
• Name-based virtual hosting
• Mixed-mode virtual hosting
• Dynamically configured mass virtual hosting

IP-Based Virtual Hosting

An IP-based virtual host allows a single server instance to be bound to a list of IP addresses and allows you to control the directories, CGI library paths, and other resources that can be accessed by the server associated with the virtual host/IP address combination. For several years, IP-based virtual hosting was popular with many ISPs that hosted Web sites. To their clients, it appeared as if each client had its own isolated Web server. If a reverse lookup on the IP address of the client's Web site were performed, it would resolve to the domain name of the Web site. With a shortage of IP addresses and serious limitations on the number of IP addresses that an ISP will assign to a shop, this form of virtual hosting is becoming limited. Today, it is impossible for an individual company to acquire their own block of public IP addresses. Your ISP generally provides as few as it possibly can. A commercial DSL connection from AT&T comes with 29 usable addresses. A few ISPs are allocating a full 256-address class-C block, but this is rare. As a result, IP-based virtual hosting is on the decline and is not as significant as it was five years ago.

For more information on IP-based virtual host, see the Apache IP-based Virtual Host Support Web site.

Name-Based Virtual Hosting

Name-based virtual hosts were introduced with the HTTP 1.1 specification and were supported through a special update to HTTP 1.0. Name-based virtual hosts exploit a new request header field in the patched HTTP 1.0 and HTTP 1.1 HOST protocol. This new parameter specifies the host name typed by the user or picked up from a hyperlink on an HTML page. If you typed www.w3c.org on your browser address bar, an HTTP 1.1 or HTTP 1.0 browser will send the following message to the server:

GET /pub/WWW/ HTTP/1.1
Host: www.w3.org

The host parameter is passed to the Apache-based server and parsed according to the rules defined in the server configuration file.

HTTP 1.1 headers are discussed in a document published on the World Wide Web Consortium's (W3C) Web site, "Hypertext Transfer Protocol - HTTP/1.1". Headers are documented in section 14.23.

Some older browsers do not support the HOST extension of HTTP 1.0 or HTTP 1.1. If an older browsers attempts to access a server that supports multiple name-based hosts, it will be given the first virtual host defined to the server. This server will become the default server.

Mixed-Mode Virtual Hosting

Unlike the original HTTP Server that supports virtual hosting using either name-based hosting or IP-based virtual hosting, the Apache-based server can support a mixture of both methods at the same time.

Dynamically Configured Mass Virtual Hosting

The cross-platform--generally UNIX--version of the Apache-based server is the choice of ISPs and companies that host Web sites because it is free. Most ISP-hosted Web sites are low activity sites and can be hosted on a single machine. The three methods of virtual hosting--IP-based, name-based, and mixed-mode--require custom configuration of the Apache configuration file and the stopping and starting of the server to implement a new virtual host. ISPs wanted a method in which new servers could be added to existing systems without requiring the stopping, starting, and configuration modification of the existing systems. The Apache Software Foundation developed a remarkable technique called dynamically configured mass virtual hosting. Once configured and implemented, you can add virtual hosts without stopping and starting your server.

For more information on dynamically configured mass virtual hosting, see Apache's Dynamically Configured Mass Virtual Hosting Web site.

Which Method Works Best with OS/400 Apache?

The following paragraphs provide step-by-step instructions to implement name-based virtual hosting. While the other methods described earlier are certainly supported by IBM's HTTP Server (powered by Apache), I am describing name-based hosting since it is the most likely choice to be used for several reasons. IP addresses are now at a premium. Few shops have IP addresses that would be required for using other methods. While dynamically configured mass virtual hosting sounds like a neat solution, there are some significant limitations regarding authentication, security, and CGI. This might be a great solution if all you are doing is hosting a large number of static page Web sites, but it's not a good solution for Web sites with e-business requirements.

Configuring a Virtual Host

Before I begin to explain the configuration of name-based virtual hosts and the directives required to implement two hosts in one server instance, I want to discuss the Domain Name System (DNS) and the DNS records necessary to support name-based virtual hosts.

DNS Considerations

Your DNS, either hosted by your company or by your ISP, needs a type-A record for each domain name that you want to translate to an IP address. If you use an NSLOOKUP tool while looking up the type-A record for the URL www.iseries.ibm.com, you will get a reply that looks something like the following:

www.iseries.ibm.com
type = CNAME, class = 1, ttl = 38097, dlen = 18
alias = as400.rochester.ibm.com as400.rochester.ibm.com
type = A, class = 1, ttl = 42289, dlen = 4
IP address = 208.222.150.11

You will see that the URL www.iseries.ibm.com is a Canonical Name (CNAME), or alias, for as400.Rochester.ibm.com on address 208.222.150.11. CNAMES, or separate domain names, point to the same IP address for virtual-named hosting. Your DNS administrator, your ISP, or you will need to implement your company's domain name (e.g., www.yourco.com) in your company's DNS. A CNAME must be defined for each alternate name within the same domain. For example, you might want to create the following URLs:

• www.yourco.com
• www.importantproduct.yourco.com
• www.other.yourco.com

You can register other domains that you might use, such as www.mybigproduct.com. Many companies register separate domains for their well-known products. Motion picture studios and television networks often create Web sites for each of their specific movies or television shows.

Figure 1 below shows my sample configuration in which I use two URLs (e.g., devtest.myco.com and devtest2.myco.com) that point to 10.130.39.223 because I did not code the UseCanonicalName off directive and I did require a CNAME for both devtest and devtest2. If I had used the UseCanonicalName off directive, I could have avoided implementing a CNAME for devtest.myco.com.

CNAME is a method used to describe how the Apache-based server constructs a self-referencing URL. A self-referencing URL is the name by which the server refers to itself, and it is required when translating abbreviated names to full names. When the UseCanonicalName on directive is used, the server will build a URL for itself as a server name and port number. When the UseCanonicalName off directive is coded, the server will use the HTTP 1.1 HOST name field and the port number to build its self-referencing URL. For more information on implementing the Apache-based server DNS limitations and constraints I recommend that you read the Apache Software Foundation's online article, "Issues Regarding Apache and DNS."

Configuring a Name-Based Virtual Host

Figure 1 illustrates a name-based virtual host configuration that supports two virtual hosts, devtest.myco.com and devtest2.myco.com. (Note: this example provided specific IP addresses; my server will start and run regardless of my DNS server's status.)

Figure 1: This HTTP Server instance contains a name-based virtual host configuration that supports two virtual hosts.

Lines 1 through 25 are identical to the single server instances that I created and described earlier. On line 3, I named the server instance using the ServerName devtest.myco.com directive. This becomes the default server for older browsers that do not support the HTTP 1.0 extension or HTTP 1.1 HOST name header field. On line 4, I am binding the server to a single IP address that is listening on port 80 using the Listen 10.130.39.223:80 directive. Lines 1 through 25 apply to the server in general and affect both of the virtual hosts that I will code.

Line 26 declares that I want to use virtual hosts on IP address 10.130.39.223. The NameVirtualHost 10.130.39.223 directive turns on virtual hosting for this server instance and enables virtual hosting for the specified IP address. I can code multiple NameVirtualHost directives if I want to support instances on multiple IP addresses.

The next important statements are the container tags. Using XML-like syntax, the first tag defines the beginning of a virtual host definition and the second terminates the container definition. Information on the lines falling within the tags controls the behavior and characteristics of the virtual host that I am defining.

On line 28, I coded the start virtual host container directive for the first virtual host as follows:



This directive tells the server that I am beginning to define a virtual host and that I want the server to listen for messages on 10.130.39.223. The cross-platform version of Apache allows you to code the wildcard asterisk (*) symbol, which is not supported in IBM's OS/400 implementation. If you were allowed to code the asterisk, the statement would instruct the server to listen for this virtual host on the IP address defined in the Listen directive as specified in line 4. An alternative is to code this directive as , which instructs the server to perform a DNS lookup to resolve the address.

The first DocumentRoot directive is coded on line 29 and it appears as follows:

DocumentRoot /www/devtest/htdocs

The next DocumentRoot directive appears on line 56 and appears as follows:

DocumentRoot /www/devtest2/htdocs

Using these directive, I am defining a specific Document root path for each of my virtual hosts. The directive on line 56 is in the second virtual host container (devtest2.myco.com), and it uses a document root directive that is different from the document root directive that is listed in line 29 (which was inserted for the first virtual host container, devtest.myco.com).

I want to serve different Web pages for each virtual server so I added the following directives to the first virtual server, devtest.myco.com. Remember, all of that server's directives are defined between the and tags on lines 28 and 53.

On line 30, I coded a ServerName directive that provides the domain name for this specific virtual host.

On lines 33 through 38, I have coded a container to define the document root directory for this server.

On lines 40 through 51, I have defined a private authenticated directory that is protected by OS/400 User Profiles. See "Putting OS/400 Apache to Work with CGI, Authentication, and SSI" for a description on how this authentication works.

The second virtual server for devtest2.myco.com begins on line 55 with a second tag and ends on line 67 with its matching tag. Line 56 defines a unique document root directory for this server. Similar to line 30, Line 57 names this server devtest2.myco.com. Lines 60 through 65 define the document directory used by this server.

Lines 69 through 97 are identical to items that I described in "Putting OS/400 Apache to Work with CGI, Authentication, and SSI," and they are available to both virtual servers, since they are coded outside the and containers.

Any directives that are not coded within a virtual host container apply to all virtual hosts. This means that both servers can call CGI programs (lines 90 through 97) or Net.Data macros (lines 81 through 88) from the same libraries. All of these directives can be moved into the virtual host containers and be made unique for each virtual host. The IBM HTTP Server (powered by Apache) gives you incredible flexibility in this area.

If you want to know which directives can and cannot be coded inside a virtual host container, refer to the Apache Directives Web site. You may also use the "HTTP Server (powered by Apache) Directives Organized by Module" online reference manual I listed earlier. IBM's online reference manual follows the Apache Software Foundation's convention for server directive documentation. For example, the following is a small segment copied from the Apache Software Foundation's documentation. IBM's OS/400 Apache documentation also follows this format:

:
directive
Syntax: directory> ...
Context: server config, virtual host
Status: Core

In this case, the and tags are used to enclose a group of directives that will apply only to the named directory and subdirectories of that directory. Any directive that is allowed in a directory context may be used.

Pay particular attention to the context line, which tells you that this directive may appear in the general server configuration or in a virtual host. For more information on virtual host matching, see Apache's online documentation, "An In-depth Discussion of Virtual Host Matching."

The Apache-based server provides an amazing array of flexible features and options to build an HTTP server that best suits your specific business requirements. Unlike the original IBM HTTP Server that supported both virtual IP and virtual name-based hosting, the Apache-based server provides many more options, allowing you to mix and match to create the exact environment that best suits your needs.

Earlier, I described virtual name-based virtual hosting in detail, but you may find that a mixture of all four methods works best for you. Still, you may want to run some single-instance servers to have better isolation and control over that single server. You also need to consider change control and the impact of running development quality assurance servers in addition to your production server on the same or different machines.

Bob Cancilla has been actively involved in the development of e-business systems using iSeries technology since 1994 and is the managing director of IGNITe/400, a nonprofit iSeries e-business user group. You can reach Bob at This email address is being protected from spambots. You need JavaScript enabled to view it..

Bob Cancilla is the IBM Rational System i Software evangelist helping to set strategy and adoption of IBM Rational application development and life cycle management software for System i customers. Bob joined IBM after over 30 years as an IT executive in the insurance industry. He was the founder of the System i eBusiness electronic user group www.ignite400.org, is the author of four books, and is an industry leader in the areas of application architecture, methodology, and large-scale integrated systems development.

 

MC Press books written by Bob Cancilla available now on the MC Press Bookstore.

 

		Getting Down to e-business with AS/400 Explains the major issues, concepts, and technologies necessary to implement an AS/400-based e-business solution—from planning for e-business to selecting an ISP. List Price $89.00 Now On Sale

Also Read

RPG Academy - Database Modernization: Entity Relationship Diagrams

RPG Academy - Database Modernization: A Bit of Database Theory

Talking About Modernization Isn’t Good Enough