Tips & Techniques > Systems Management

AIX Health Check

Understanding available resources, automation and configuration of system-administration functions

Systems management means different things to different people. It requires various utilities and commands to help with many systems-administration functions, including making operational configuration changes, systems monitoring and system documentation.

Administrators use many methods to adhere to these concepts, such as the AIX* Systems Management Interface Tool (SMIT), WebSM or the command line. SMIT comes in two operation modes, giving users fast paths to the most frequently used administrative tasks. Newcomers to AIX may not know that SMIT is often preferred over the command line for making changes for many operations. For example, if you use ifconfig to make a routing change, it won't save the information after a reboot. Instead, consider the Object Data Manager, a critical link of AIX.

Regarding partitioning information, we used to be limited to using a hardware-management console (HMC); now users can choose the Integrated Virtualization Manager, which lets them manage LPARs and configure virtual Ethernet and SCSI without an HMC. You can also back up and restore LPAR configuration information and view application logs and device inventory, particularly important for the documentation section.

Monitoring With nmon

One of the best utilities I've seen for AIX, interestingly, isn't even a supported IBM* utility, but nmon. The beauty of nmon is that you can use it either to quickly ascertain what's going on in your system or to capture data for historical trending and analysis. Most tools do one or the other, but nmon does both equally well.

Using nmon to capture and examine data is simple. After you've finished, you must sort the file and rename it with the .csv suffix, then FTP the file to your PC, start what's called the nmon analyzer and click on "Analyze nmon Data." You should see the output on a spreadsheet with graphs, shown in Figure 1. The only thing that nmon lacks is the capability to present this data on many LPARs concurrently, though you can use a tool called Ganglia to integrate the nmon analysis into a database.

The lparstat command is also useful. It provides information regarding performance and systems configuration, such as mode, entitled capacity and if it's an uncapped partition. A similar command is mpstat, which displays the overall performance of all logical CPUs on the systems. You can run smtctl to confirm that you're running in simultaneous multithreading (SMT) mode.

Another wonderful command I use frequently - particularly when I'm asked to work on a box I'm unfamiliar with - is prtconf. This command gives you partitioned hardware information (CPU, clock speed, RAM, firmware level, devices), network information, paging space, volume group information, etc.

I'm a big believer in trying to get as much information as possible before doing anything to a system, including monitoring and systems tuning. How can you monitor a system - let alone tune it - without taking the time to understand how it's built? Several commands, such as lsattr, can be run to get kernel information about your CPU, disk and networking parameters.

Next page: >>