You have to kill -9 them, when you let run a jstack on a JVM pid you'll see WAITING threads.
Best, Alex On Dec 19, 2012, at 12:34 PM, Shengjie Min <[email protected]> wrote: > Hi, > > I noticed that when my flume master is down for a quite while. My flume > agent is retrying to connect to the master, i can tell that from the flume > logs as below; > > 2012-12-19 11:11:42,126 WARN com.cloudera.flume.agent.MultiMasterRPC: Could > not connect to any master nodes (tried 1: [flume-master-01:35872]) > 2012-12-19 11:11:52,140 WARN com.cloudera.flume.agent.MultiMasterRPC: Could > not connect to any master nodes (tried 1: [flume-master-01:35872]) > 2012-12-19 11:12:02,152 WARN com.cloudera.flume.agent.MultiMasterRPC: Could > not connect to any master nodes (tried 1: [flume-master-01:35872]) > 2012-12-19 11:12:12,162 WARN com.cloudera.flume.agent.MultiMasterRPC: Could > not connect to any master nodes (tried 1: [flume-master-01:35872]) > > from "top", I've seen a lot of java processes owned by flume. followed my > instinct, I stopped flume agent, after a while, I run "top" on it, I still > have a lot of processes owned by flume user. > > 12735 flume 20 0 1487m 26m 128 S 0.3 1.3 0:05.41 java > > > > > 13828 flume 20 0 1487m 24 24 S 0.3 0.0 8:11.92 java > > > > > 14442 flume 20 0 1487m 0 0 S 0.3 0.0 8:06.81 java > > > > > 15033 flume 20 0 1487m 26m 164 S 0.3 1.3 0:04.29 java > > > > > 15453 flume 20 0 1487m 4 4 S 0.3 0.0 8:07.55 java > > > > > 15964 flume 20 0 1487m 27m 164 S 0.3 1.4 0:04.09 java > > > > > 16098 flume 20 0 1487m 0 0 S 0.3 0.0 8:12.67 java > ...................... > > Why so many processes are generated here and they don't go away even when > flume is stopped. > > -- > All the best, > Shengjie Min
