Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Cannot Reproduce
-
None
-
None
-
None
-
Windows
Hadoop 2.5.0cdh5.2.0
Accumulo 1.6.0cdh5.1.0
Description
When I run the MiniAccumuloCluster 1.6.0 on Windows, it does not start up correctly, but ends up with the error
java.lang.RuntimeException: Zookeeper did not start within 20 seconds. Check the logs in C:\Users\jsbaue\AppData\Local\Temp\1415364318664-0\logs for errors. Last exception: java.net.ConnectException: Connection refused: connect
at org.apache.accumulo.minicluster.impl.MiniAccumuloClusterImpl.start(MiniAccumuloClusterImpl.java:460)
at org.apache.accumulo.minicluster.MiniAccumuloCluster.start(MiniAccumuloCluster.java:72)
at com.gfk.st2.pace.datafactory.adaptor.accumulo.impl.MiniAccumuloClusterTest.testStart(MiniAccumuloClusterTest.java:50)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
Reason is, that Zookeeper does not startup at all, as the classpath, returned by the function
private String getClasspath() throws IOException
in MiniAccumuloClusterImpl.java returns a classpath, containing just the config directory. The reason for this is the part
// assume 0 is the system classloader and skip it for (int i = 1; i < classloaders.size(); i++) {
(Line 208f in 1.6.0, 224f in 1.6.1), my debug-situation can be seen in screenshot-1.png. There you also can see, that the list classloaders includes only one classloader, which is a java.net.URLClassLoader. This classloader does not have a parent. Subsequently, the list has only one element, and the for-loop is never entered. So the classpath does not even contain the class org.apache.zookeeper.server.ZooKeeperServerMain, so Zeekeeper cannot be started.
When I start the loop with 0, zookeeper starts up and I can use the MiniAccumuloCluster.
Attachments
Attachments
Issue Links
- is related to
-
ACCUMULO-3380 MiniAccumuloCluster does not work with long classpath under Windows.
- Resolved