Hardware installation is the part of system administration that makes it different from every other white-collar job. Although most of your time is spent in front of a keyboard or monitor, you often step out of the office and into the data center where you can perform heavy lifting, work with power tools, and dirty your clothes with dust and grease. Many administrators actually find this relaxing, helping them to get away from the daily grind of the office.
Most people are tempted to open shipments as soon as they arrive and begin racking their new hardware. This haphazard approach to deployment can result in many logistical problems, like running out of nearby network ports or running out of room in a rack. You should have some general procedures and guidelines for each installation you perform; a generic installation might include the following tasks:
Transport hardware to your data center.
If necessary, build and install racks or cabinets.
Map the placement of each item on your racks.
Remove hardware from its containers.
Label the hardware.
If necessary, install mounting hardware.
Install any third-party components.
Install the hardware in your racks using the mounting hardware.
Attach cables and adapters.
Power on hardware and verify its functionality.
If you receive hardware at one location and must ship it yourself to another location for installation, take extreme care to transport the equipment safely. Keep everything in its original boxes and secured by the original packing material. Securely tie or strap the equipment in the transport vehicle so that it can't shift or slide during the move. Rarely can you move data center equipment in a carmost cars aren't large enough to handle the boxes. Consider renting a van or truck specifically for the move. Even if you have a personal truck or SUV large enough to transport the equipment, you should consider whether or not your liability insurance is adequate to cover the value of the hardware you're carrying. Transportation of company property in personal vehicles rarely is insured for as much as property transported in a company-owned vehicle. This means that if your company is covered for only $100,000 of property transported in personal vehicles, that's all the equipment you should take with you in any one trip.
Anyone can open and unpack a box, but some care should be taken with computing hardware. The most important rule to follow is to never unpack the contents of a box until it reaches its final destination. Manufacturers ship hardware in such a way as to minimize the impact to it if the box shakes, falls, or is otherwise damaged. If you remove a server from a box before transporting it in your car, simply running over a pothole could result in serious damage.
When you finally do unpack your boxes, do not throw anything out! This includes plastic bags, bubble wrap, Styrofoam, and the boxes themselvesall things that you might otherwise be inclined to throw out. If you need to ship something back or move it again, you will need all of these things to repack and properly transport the hardware.
You may not have room to store all of this material in your office or data center. The best way to save space in this situation is to break down all of the boxes and flatten them. If you have no room for the packing material, you can throw it out after you are sure the hardware works. If you ever need to repack it, you can purchase bubble wrap or Styrofoam peanuts from most office stores and shipping companies.
Unpacking Heavy Equipment
Heavy equipment like monitors and servers are very difficult to lift out of a box, especially when flanked by several inches of Styrofoam. Instead of throwing your back out or risking dropping the hardware, turn the box on its side and cut the bottom flaps open with a box cutter. While holding the contents in place with one hand, gently tip the box so the bottom flaps are on the floor facing outward; the Styrofoam should keep the hardware from falling out. Now you can gradually lift the box off of the hardware, which is a much easier proposition than lifting the actual hardware.
After you have removed hardware from its packaging, label it immediately. Each item should be identified as soon as it leaves the box in order to prevent confusion when it comes time to mount the hardware in racks. This is especially true when you have many similar-looking machinesfor example, it is a lot easier to find a machine labeled db2 than it is to find the Sun Enterprise 420R with 4GB of memory instead of 2GB in a sea of other Sun Enterprise 420R servers.
Use a Label Maker
Purchase a label maker before starting any deployment and keep it at your data center. Use labels with a solid background (white or black) so you can read them on hardware of all colors.
Mounting hardware in racksfrequently referred to as racking hardwarecan be the most excruciating part of any data center installation. You're constantly lifting and balancing heavy hardware, reaching for inaccessible screws, fitting greasy sliding rails, and generally getting your physical exercise for the day. Following a few simple guidelines can help make the job as manageable as possible and ensure that you get the installation right the first time.
Unless you are building your own data center, your hosting facility will most likely provide racks or cabinets for you. Many racks require some assembly, eliminating the time-consuming hassle of building them yourself.
You need at least one other person to help you with the installation, so make sure that person is ready and onsite before you begin. You'll need your assistant to help unpack the material, as you learned in the preceding section, but you'll also need that person to hold heavy equipment while you're mounting it in the rack. Trying to balance a server with one hand and tighten screws with the other is just asking for trouble. Use an assistant to prevent injury to both you and the equipment, and to make the installation process as quick and efficient as possible.
Individual pieces of hardware fall into one of four "racking" types:
Hardware with sliding mounting rails
Hardware with mounting brackets
Hardware with screw holes for direct mounting onto a rack
Hardware that cannot be racked
Sliding rails were introduced in Chapter 2, "Designing the Data Center Infrastructure"; they allow you to slide servers in and out of the rack for maintenance without actually removing it from the rack itself. Most sliding rail systems are equipped with a safety locking mechanism that prevents you from pulling a server completely out of the rails without pushing a button or lever.
Sliding rail systems usually require installation of rails on both the servers and the racks; two people are needed to properly insert heavier hardware into the rails. These systems also require a four-post rack or a cabinet to support the weight of a server in various positions. Figure 3.2 shows how a server on sliding rails can be easily pulled out for maintenance.
Figure 3.2 Sun Enterprise 420R servers mounted on sliding rails in a cabinet. The bottom server is pulled out for maintenance without actually removing it from the rails.
Mounting brackets are directly attached to a rack and only hold hardware in place once it is in its proper mounted position. Unlike sliding rails, mounting brackets won't support a piece of hardware as you are inserting it into or pulling it out of a rack. You can use mounting brackets on two- or four-post racks.
Some hardware just comes with screw holes on the chassis itself; you can use these screw holes to mount the hardware directly onto any kind of rack. Be careful when working with this kind of hardwareyou need at least one other person to support it from the bottom while you are installing or removing it.
Finally, there is hardware that cannot be racked. Few servers these days fall into this category, but many peripherals, such as small tape libraries and disk arrays, do not support rack mounting. You may also have a server that is incompatible with your rack configuration (such as with sliding rails on a two-post rack). You can place each of these items on a shelf (as discussed in Chapter 2). Still other hardware, such as large tape libraries and high-end servers like the Sun Enterprise 10000, are simply too large to fit into a rack and are meant to be placed on the floor.
With all of this in mind, you should draw a map of where the hardware should be mounted in your racks, paying special attention to the height of each piece of hardware and the location of any vents. Both rack and hardware height is measured in rack units (U), which are equivalent to 1.75 inches. This measurement makes it very easy to calculate how many pieces of hardware you can fit in a rack. For example, a Sun Enterprise 420R is 4U in height; you know you could fit 10 of them in a 41U cabinet with 1U to spare. A Sun Netra t1 105 is 1U in height; you could fit 41 of them in the same space.
Using Rack Units
Always refer to hardware height in rack units. Working with whole numbers is much easier than working with fractions of an inch. Although most vendors list the total height of a rack in inches, the usable height is usually given in rack units.
Begin mounting your hardware at the bottom of a rack, and move your way up to the top; that way, you can use a racked server as a temporary support for the next one, if your arms get tired later in the process.
Place hardware that can't be racked on mountable shelves. Space the shelves so that the hardware they support can get the proper ventilation. You also should space the shelves so that you can insert and remove them without interfering with nearby hardware. To avoid all possibility of a "space crunch" near shelves, you can put all of your shelved equipment at the top of the rack so as not to interfere with any mounted hardware.
Verify Shelf Load Capacities
Shelves should be able to handle the weight of the hardware they support. Always request the specs of the shelves from your vendor and make sure you buy shelves that can handle the equipment you need to place upon them.
If you are installing a console with a keyboard and monitor, place them at a comfortable working height for the average person on your staff. You can place keyboards and monitors on a shelf, or you may have a KVM (keyboard, video, mouse) switch with a keyboard and monitor in one mountable module, as is shown in Figure 3.3. Either way, remember that most users stand when working at this type of console, so a working height somewhere between 3 and 4 feet above the floor is appropriate.
Figure 3.3 A rack-mounted KVM switch with an integrated monitor and keyboard.
A keyboard, video, and mouse (KVM) switch is a very convenient piece of hardware. It allows you to use a single keyboard, mouse, and monitor on multiple machines, saving space and money. It has inputs for the keyboard, video, and mouse connections from each of your servers and separate output ports for the shared keyboard, mouse, and monitor. A physical switch or keystroke combination switches the active server, eliminating the need to run back and forth swapping cables just to access consoles on two different servers.
Real-World Example: The Server Bench Press
One system administrator recruited a coworker to assist him with mounting six 65-pound Sun E420R servers in a rack. This turned out to be a smart move. After mounting two of the servers, one administrator tried to insert the third server into its sliding rails. Unfortunately, the server got stuck because the rails were mounted unevenly; one rail was mounted one hole below the other. It's very easy for this to happen, but he should have checked the rails before trying to mount the server. The unlucky administrator tried to pull the server off of the rails, but it was stuck in the rails, and the rails were fully extended. The administrator couldn't let go of the server or it would have crashed to the floor. Luckily, his assistant was available to hold the heavy server until the administrator could pry it from the rails. The administrator then remounted the rails and installed the server. This incident illustrates the necessity of having at least one person assist with heavy equipment installations. Don't go it alone.
Attaching Cables and Adapters
After hardware is mounted, you can begin plugging in various cables and adapters to attach it to power sources, network devices, disk arrays, and any other equipment you may have. You should tighten any screws that fasten adapters to the ports on your hardware; an insecure connection on any port could result in an unreliable signal causing random data loss and errors.
You should be especially careful when plugging in cables and adapters with very thin pins, like those on most SCSI cables. Even the slightest offset from the hole can cause a pin to bend or break, rendering the cable useless. You may be able to fix bent pins, but you run the risk of worsening the problem by breaking it or bending other pins.
You will also want to use this time to label your cables. Each label should indicate the type of cable and the hardware at either end of the connection, including interface names. For example, a category 5 cable connecting the eth0 network interface on a Linux server called apple to a 100Mb switch might read:
apple eth0/100 Mb Ethernet
Real-World Example: Using Colored Cables
Some organizations use colored cables to indicate their purpose. One such company used category 6 cable for its network connectivity, digital phone lines, and analog phone lines. The company used red, white, and blue cables to differentiate each function so administrators could clearly see what kind of traffic was flowing over each cable.
Premature Power On
Make sure the power switch is turned off before plugging in a power cable; otherwise, the machine may power on before you are ready.
Special care should be taken with power cables. As was mentioned in Chapter 2, you should separate power cables from other types of cables; the electromagnetic interference from power cables can introduce errors into the signals on other cables.
Testing Your Hardware
After your hardware has been racked and connected to other devices, you should test to see that everything powers on correctly. For most systems, this is as simple as flipping the switch and verifying that any fans or LED indicators are activated; you haven't installed any software yet, so there is not much else you can test for at this point.
You should test other peripheral hardware for basic functionality. For instance, a tape library, without any supporting software, should be able to accept a tape in one of its bays, scan the tape's barcode label, and recognize it, perhaps showing the label on a display. Some types of hardware actually perform POST (Power On Self Test) operations, which test most of the internal mechanisms and display the results on an LCD display or by flashing LEDs. All hardware should pass some basic functionality test; your documentation should tell you what to expect when you power on a system and how to test it.