A high load process analysis, very valuable (turn)

Analysis of the problem of US high and Sy High:

When the US value is too high, it means that the running app consumes a lot of CPU. The reason why Java applications cause us high is that threads are always in a running (Runnable) state, usually caused by tasks such as non-blocking, looping, regular, or purely computational, and the other one that may also cause US high is frequent GC.

When the SY value is high, it means that Linux spends more time on Java thread switching. The main cause of this behavior in Java applications is that more threads are being started, and most of these threads are constantly blocking (for example, lock waits, io wait states) and changes in the execution state, which causes the operating system to constantly switch execution threads, resulting in a large number of thread context switches.

For Jstack to do the threaddump stack, can reflect the following information.

If an identical call stack appears frequently, we have more than 80% reasons to determine that the code has a performance problem (except for the portion of the Read network);

If the same call stack appears on the same thread (TID), we have great reason to believe that there may be more loops or loops in this code;

If a call stack appears frequently and has a lock inside, check the cause of this lock, which may be a performance problem with global lock;

In a less stressful cluster (w<2), we rarely get stacks with business code, and typically in a full stack, up to 1-2 of the business code stack,

If this occurs frequently, be sure to check the code for performance issues.

If you suspect a dead lock problem, then please find out all the lock IDs to see if there is a duplicate lock ID.

http://blog.csdn.net/axman/article/details/7104819

A high load process analysis, very valuable (turn)