this link:
http://blog.csdn.net/gklifg/article/details/50418109、http://blog.csdn.net/levy_cui/article/details/53992402、http://blog.csdn.net/lulynn/article/ Details / 46455097。
In the process of using the CDH cluster, the problem is that Hiveserver2 has just started and the client has abnormally terminated after visiting the client.
1. Process Analysis:
See the above three blogs, and use JSTAT to monitor. The following prompt will appear:
When the Hiveserver2 red box was killed;
During CDH monitoring, this is called an Outefmemory:
2. Temporarily resolved:
The prompt appears here, indicating that the memory is not enough, the memory is increasing, and the tuning content is as follows:
The initial configuration is as follows:
Currently use JSTAT Monitor and you can find:
You can see from the data above:
1. The memory will be swapped between S0 and S1. When Hiveserver2 exits for the first time, it finds that S0 and S1 are 100%. At the same time, full GC is executed multiple times to cause memory overflow;
2. When the parameters are modified, the new generation GC is not often found. After each YGC, the utilization share in the E region will decrease;
3. Increasing the area of occupation. When full GC occurs, the O region can be cleared, so the fraction will decrease;
Share, grow, be happy
Popular and down to earth
Please enter the blog address: http://blog.csdn.net/fansy1990