Guest post by Ramakrishna Baipadithaya Kenchabhatre and Sunil Lingappa, HP Operations Agent R&D Leads
Meet Jim. Jim is an IT administrator at a large enterprise company which has a heterogeneous mix of standalone and virtualized systems. They have recently implemented a virtualized-based solution for a section of the enterprise which deals with test automation using AIX Logical Partitions.
Jim is flooded with complaints from users of the test systems. They complain that the virtual machines are too slow to respond. Jim who is an expert in debugging standalone system problems invokes his favorite tool glance (part of GlancePlus package). Glance can now be used to debug both standalone and virtualized environments. It helps him quickly troubleshoot the problem, identify the root cause and rectify it across different virtualization technologies like VMWare, IBM LPARs etc. Jim now gets down to the work at hand.
Jim is aware that the resources are shared and the troubleshooting cannot be done in a silo. The decrease in speed could be due to the fact that the neighboring Virtual Machines (VM) are consuming more resources than expected. As the resources are shared, the target VM is penalized because of over-resource usage by other VMs. Jim invokes ‘glance -V’ from a console to troubleshoot the problem and sees the following screen.
Figure 1: Glance screen showing list of logical systems
The screen nicely shows the list of logical systems along with the Logical CPU, Physical CPU and memory utilizations. Because he now has a holistic view of all the virtual machines and the hosts, it is now easier to see the resource utilization by different logical systems. He can do a further drill down on each system if needed and the detailed screen looks as below.
Figure 2: Glance drill down screen for a logical system
If multiple VMs are having the resource utilization issue, Jim checks if they belong to the same host. If they do, then the hosts’ utilization needs to be scrutinized. If it is only a single VM, Jim drills down to that VM to see how the resources are getting utilized. To have further details, he can login to the respective system and run ‘glance’ on it.
Using this scenario we see that GlancePlus provides a “right now” perspective of performance metrics to troubleshoot system performance bottlenecks in a realistic manner. It provides one consistent user interface across different virtualized environments and minimizing the learning curve.