Managing RHEL Systems

The ability to manage and grow your Linux infrastructure efficiently (and inexpensively) are probably the best reasons for using RHEL systems. Red Hat created the Red Hat Network to allow you to track and update your RHEL systems properly. With different subscription levels come increased levels of support and features for maintaining anywhere from one to thousands of systems.

Using features including global file systems (GFS) and high performance computing (HPC) clustering techniques, Red Hat offers help creating a flexible and efficient Linux software infrastructure. This infrastructure lets you expand your computing capacity incrementally without discarding (or even stopping) the structure you put in place.

Using Red Hat Network

The Red Hat Network ( https ://rhn.redhat.com ) offers a service for managing your Linux systems. You can try out Red Hat Network when you sign up for an evaluation subscription. From the Red Hat Network site, you can upgrade your demo system to a paid subscription or sign up for additional entitlements to manage multiple RHEL systems.

Red Hat developed Red Hat Network to help companies implement and manage their Linux infrastructure. There are several different architectures in which Red Hat Network can be configured. These different architectures allow you to choose whether or not your RHEL systems talk directly to Red Hat Network (to manage system entitlements and updates), or to a satellite or proxy server set up on your premises.

The architecture you choose will impact the level of control you can exert over your RHEL systems. Here are the choices:

• Hosted model - With a hosted model, each RHEL system communicates directly with Red Hat Network servers for managing the updates it gets. Those updates are based on the appropriate channel for the hardware and software you are using. That information is stored by Red Hat servers, as opposed to being managed locally on your own network.

• Proxy model - The proxy model works similarly to the hosted model, but instead of having each system on your premises get updates individually from Red Hat Network, a proxy server is set up on the customer premises, so each set of updates needs to be downloaded to your network only once. Individual systems will then get the updates they need from the local proxy server. Although updates are stored locally, information about channels and entitlements are still managed by Red Hat Network servers.

• Satellite model - All Red Hat Network features (system management, updates, and so on) are located on your local network. Updates are downloaded from Red Hat Network to your satellite server, as they are with a proxy server. However, features for managing custom channels and other tasks normally done by RHN can be offloaded to the satellite server. Different levels of support and training are available, although the intent is to have Red Hat act as technical consultant while your own employees manage system and database administration for your network satellite server.

The basic Update service, up2date, is available with all RHN accounts from a demo system on up. The up2date facility checks for available software updates on an on-going basis (using the rhnsd daemon) and alerts you when updates are available (via a red exclamation mark icon on the desktop). This facility makes it easy to keep up with the latest errata patches. (See Chapter 5 for a description of the up2date facility.) Other features that come with the basic service are the ability to download ISO images of RHEL installation disks, and search for and get selected individual software packages to install.

Beyond the basic Update service, there are levels of support you can purchase from Red Hat. The Management and Provisioning support programs are aimed at helping you manage RHEL systems across an entire organization. Here's what you get with those services:

• Management - With this subscription level, an organization can set up administrative users to manage groups of systems. An Organization Administrator has the ability to assign those users and the systems to which they have access. To manage their systems, a user can do such things as compare package sets installed on different systems, search groups of systems for information on installed packages or hardware tags, group systems together so they can be managed as a group , and process adding or removing software packages for a batch of systems.

• Provisioning - The Provisioning subscription level includes all features of the Update and Management levels, but adds further features for managing multiple systems. For example, you can create kickstart profiles that allow you to install groups of systems with many features preset (such as time zone, GPG and SSL keys, and specific package sets). If you are installing on similar hardware, you can even preset normally tricky configuration features such as video card configuration. Provisioning subscribers also have the ability to roll back their system software configurations to points in the past. Using snapshots when updates are done, you can go back to previous configuration files, package profiles, and channels.

Besides software management and update features, Red Hat Network offers features for monitoring your registered systems. This can include simple queries ( pings ) to make sure systems are up and running, as well as responses to alert you to dangerous system states (such as WARNING and CRITICAL log messages). Reports can be generated to provide overviews of your system states.

To find out more about how Red Hat Network works and the features it has available, refer to the Red Hat Network Reference Guide ( http://rhn.redhat.com/help/reference/ ).

Using RHEL for High-Performance Clusters

Cluster-based High-Performance Computing (HPC) begins with the concept that you don't need huge, monolithic computers to handle high-demand business and scientific applications. Instead, you can simply put a structure in place that lets you add inexpensive, commodity hardware and open source software to get to the capacity level you need.

Open source software projects such as Ganglia ( http://ganglia. sourceforge .net ) and Platform Rocks ( www.platform.com/products/rocks ) have emerged in recent years to help form the foundation for managing and deploying cluster computing systems. Red Hat offers subscriptions aimed directly at supporting HPC clusters. Contact Red Hat directly for more information.

With HPC clusters, computing that needs to be done within an organization can be spread across multiple compute nodes (also called worker nodes ). All compute nodes are configured identically. Master nodes are set up to direct the work of the compute nodes. Compute nodes run RHEL WS systems, while the master nodes must be an ES or AS system.

Using RHEL Global File System

For organizations with large-scale demands on Web servers, mail servers, file servers, or database servers, Red Hat Global File System (GFS) offers the capability to centralize data storage in a way that is efficient and scalable. The idea is to have all servers that need to access a company's data refer to one namespace that would be lightning-fast to access.

With GFS, RHEL servers are connected to a storage area network (SAN) that acts as a cluster file system. The fact that a GFS file system may reside on multiple disks on multiple storage units is invisible to the people and applications using the data.

Because GFS is a file system type and volume manager, by supporting GFS all Red Hat servers can access GFS volumes as they would any file system. Like file system types such as NFS, GFS can be accessed over a network. Unlike NFS, however, GFS doesn't incur the overhead of using a TCP/IP-based network for communication, but instead relies on iSCSI or Fibre Channel for communications.

Because communication can be done so efficiently using GFS, there is no need to incur the complexity of maintaining multiple versions of data to ensure good performance. Any component of a GFS configuration (storage units, servers, or network interfaces) can be redundant. So any component that goes down can be quickly replaced . If increased capacity is needed, you can add volumes or resize file systems while the GFS is still up.

Because growth of your GFS can be done by scaling out (adding machines of the same architecture) instead of scaling up (replacing existing machines with better ones), businesses can grow without taking down existing infrastructure. Expansion can be done incrementally.