the result with skywalking-agent-6.0.0-alpha-adjust-agent-setting. 
jdk version is:  It seems the method to validate is not correct. I read the response from @raphw again, considered with the opportunity of actual "leak". In my understanding, Its not the real leak as generally speaking. Its occur just because huge native call when transform jar files from local disc, and remaining os mem is not enough to continue before app recycled them. It seems the correct method contains two dimensions. 1. Replay the actual scenes(weibocom/motan#756) with using skywalking-agent 3.2.6 and 6.0.0-alpha-adjust-agent-setting. If it is not killed when using skywalking-6.0.0-alpha, it will be proved worked. "weibocom/motan#756" in english, its main idea is that when jvm heap and remaining os mem is small, the os/kubernetes will kill the "jvm instance"/"docker instance". 2.In the machine that owns enough mem to run gperftools to analysis the result again, compare with below:  If the key of result below is much less than above, it will be proved worked. like Java_java_util_zip_xxx   Am I right about above two dimensions to validate? @raphw @wu-sheng . [ Full content available at: https://github.com/apache/incubator-skywalking/issues/1666 ] This message was relayed via gitbox.apache.org for [email protected]
