Tchater maintenant avec le support
Tchattez avec un ingénieur du support

Foglight 7.1.0 - Performance Tuning Field Guide

Overview Hardware and Operating System Tuning Management Server Tuning Java Virtual Machine Tuning Backend Database Tuning High Availability (HA) Tuning Agent Tuning Store and Forward Processing Appendix: Performance Health Check Appendix: Analyzing a Support Bundle

CPU

Before you run the Foglight® Management Server in a virtual image, you must make sure that the planned number of processors will be made available. In certain scenarios, the virtual image will not exist exactly as planned.

To confirm how many processors the virtual image, and therefore the Management Server, will receive, generate a support bundle and check the Diagnostic Snapshot. In the Diagnostic Snapshot, look for AvailableProcessors.

Ready Time

It is important to monitor the Percent Ready value for each VM. A Percent Ready value in excess of zero (0) indicates that, of the amount of time the process was ready to run, for that percentage of the time it could not because it was waiting for resources. A Percent Ready value under two percent (2%) is considered good; values in excess of 2% can pose a challenge.

When a Percent Ready value is in excess of 2% for a few seconds or longer, the application might become jerky and appear to pause. This is why the responsiveness of an application might deteriorate after a certain number of users is reached.

Since ready time is the time a process spends waiting when it could be running, it has a direct impact on response time.

For example: A number of virtual machines are pinned to a single CPU and 600 seconds of CPU time are available. Four of the virtual machines have load and about 300 seconds of ready time has accumulated, so you can expect a slowdown of approximately 50 percent. An event with a response time of 1 second could be expected to take 1.5 seconds under that level of load as a direct result of the delays caused by the ready time. This is not atypical for observations of multiuser systems under load. The same factors apply to an ESX® Server host running multiple virtual machines. Also, the more virtual CPUs the VM has, the more CPUs have to be free, or the VM may have to wait longer for x number of CPUs to be free.

Memory

Before you run the Foglight® Management Server in a virtual image, make sure that the image is configured with the necessary amount of memory.

The Management Server is very active and consistently uses a significant portion of the allocated memory, so ensure that the memory you allocate to the virtual image is committed. For Management Server performance, it is far better to assign the appropriate amount of memory and configure it to be fixed than to assign more memory and configure it to be dynamic.

Also, it is often the case that plenty memory has been allocated to the VM in question, but the memory limit has been left at a level that is too low for the VM to obtain the memory it needs.

To confirm how much memory will be allocated to the virtual image, and therefore to the Management Server, generate a support bundle and check the Diagnostic Snapshot.

Diagnostic Snapshot

---- jboss.system:type=ServerInfo
MBean attributes:
attr ActiveThreadCount = 390
attr AvailableProcessors = 1
attr OSArch = amd64
attr MaxMemory = 9556328448
attr HostAddress = 10.4.52.60
attr JavaVersion = 1.6.0_04
attr OSVersion = 2.6.9-42.EL
attr JavaVendor = Sun Microsystems Inc.
attr TotalMemory = 9556328448
attr ActiveThreadGroupCount = 21
attr OSName = Linux
attr FreeMemory = 6509179336
attr HostName = host25.example.corp
attr JavaVMVersion = 10.0-b19
attr JavaVMVendor = Sun Microsystems Inc.
attr JavaVMName = Java HotSpot(TM) 64-Bit Server VM

Review the MaxMemory, TotalMemory, and FreeMemory entries. For more information about the diagnostic snapshot, see Diagnostic Snapshot .

Documents connexes

The document was helpful.

Sélectionner une évaluation

I easily found the information I needed.

Sélectionner une évaluation