GridGain - programmatically opening nodes using SSH through Grid.startNodes API

问题

I am using Grid.startNodes(java.util.Collection, java.util.Map, boolean, int, int) as defined here: http://gridgain.com/api/javadoc/org/gridgain/grid/Grid.html#startNodes(java.util.Collection, java.util.Map, boolean, int, int)

Code I am using:

GridConfiguration cfg = GridCfgGenerator.GetConfigurations(true);
Grid grid = GridGain.start(cfg);

Collection<Map<String,Object>> coll = new ArrayList<>();

Map<String, Object> host = new HashMap<String, Object>();
//host.put("host", "23.101.201.136");
host.put("host", "10.0.0.4");
host.put("port", 22);
host.put("uname", "username");
host.put("passwd", "password");
host.put("nodes", 7);
//host.put("ggHome", null); /* don't state so that it will use GRIDGAIN_HOME enviroment var */
host.put("cfg", "/config/partitioned.xml");

coll.add(host);

GridFuture f = grid.startNodes(coll, null, false, 3600 * 3600, 4);
System.out.println("before f.get()");
f.get();

• I ran the above code on a vm with a 10.0.0.7
• I have remote desktop into the VM whos host IP is 10.0.0.4 and see no changes to state. The code completes and exits. Both VMs are able to run gridgain locally and can discover each other's nodes if I start it using bin/ggstart.bat
• I can manually start a node on 10.0.0.4 (the machine I am trying to SSH into via this API). I can start said node by running $GG_HOME/bin/ggstart.bat $GG_HOME/config/partitioned.xml so there is no issue in the configuration file

I am not quite sure how to debug this as I get no errors

回答1:

Successful completion of the future returned from startNodes(..) method means that your local node has established SSH session and executed a command for each node it was going to start. But successful execution of a command doesn't mean that a node will be actually started, because it can fail for several reasons (e.g., wrong GRIDGAIN_HOME).

You should check the following:

• Are there GridGain logs created GRIDGAIN_HOME/work/log directory? If yes, then check them - there could be an exception during startup process.
• If there are no new logs, there is something wrong with the executed command. The command can be found in the local node logs - search for "Starting remote node with SSH command: ..." lines. You can try to create an SSH connection in terminal, run this command and see what happens.
• Also you may want to check your SSH logs to see whether there are any errors.

来源：https://stackoverflow.com/questions/26188446/gridgain-programmatically-opening-nodes-using-ssh-through-grid-startnodes-api