JPPF, java, parallel computing, distributed computing, grid computing, parallel, distributed, cluster, grid, cloud, open source, android, .net

The open source
grid computing

 Home   About   Features   Download   Documentation   Forums 
September 20, 2017, 04:15:24 AM *
Please login or register.

Login with username, password and session length
Advanced search  
News: New users, please read this message. Thank you!
  Home Help Search Login Register  

Recent Posts

Pages: [1] 2 3 ... 10
Troubleshooting / Executing tasks from different Jobs in the same node
« Last post by lukss12 on August 25, 2017, 04:49:27 AM »
I'm very new with JPPF and I'm trying to use it for a college research on distributed execution models.
I have a little question I'm sure you can answer:

-First suppose this setup: A machine running a driver instance and a node instance the same time.

I run a JPPF client configured programatically to use 4 connections to the driver so I can send 4 Jobs to be executed the same time.
This is done correctly because I can see "Executing" state for the 4 jobs in the admin-ui.
If I put 1 task in every Job then my driver executes the tasks one by one but if I put 4 tasks in every Job the four tasks are executed together - that is telling me that my driver is configured properly to send up to 4 tasks to a node -
So I wonder if there is a way to tell the driver to execute up to 4 tasks from different Jobs in the same node simultaneously for the cases where a job has less than 4 tasks.

Thank you very much and sorry for my English  :-\
Installation and Configuration / Execution channell count
« Last post by arefaydi on August 23, 2017, 08:42:39 AM »
Is there a way to get count of nodes and peer driver connected to driver (or execution channell count) programmatically. I want to divide jobs to tasks according to execution channell count at that moment.
Installation and Configuration / Re: Grid topology
« Last post by aburak on August 18, 2017, 10:09:48 AM »
The solution worked, thanks a lot  :D.
Installation and Configuration / Re: Grid topology
« Last post by lolo on August 18, 2017, 05:12:37 AM »

To prevent two drivers from seeing each other, you must disable the automatic peer driver discovery mechanism, which is enabled by default. This is done in the driver's configuration with this property:
Code: [Select]
jppf.peer.discovery.enabled = false
Similarly, if you wish to have two distinct groups of nodes, each seeing only one of the drivers, the simplest is to:

- disable the drivers broadcasting service with this configuration property:
Code: [Select]
jppf.discovery.enabled = false
- disable each node's driver discovery service and configure the connection to the driver manually:
Code: [Select]
# disable driver discovery
jppf.discovery.enabled = false
# connect to driver on specified host and port = driver1_host
jppf.server.port = 11111

In the same way, you may disable driver discovery in your clients and manually configure which driver(s) they connect to:
Code: [Select]
# disable driver discovery
jppf.discovery.enabled = false
# names of the drivers to connect to
jppf.drivers = driver1
# connect to the driver on specified host and port = driver1_host
driver1.jppf.server.port = 11111

With this approach, you can setup two or more distinct JPPF grids on the same network, each grid being compeltely independant from the others. You can even co-locate multiple drivers on the same host, provided you assign each of them a separate port with jppf.server.port = <port_number>

I hope this clarifies,
Installation and Configuration / Grid topology
« Last post by aburak on August 17, 2017, 01:15:03 PM »

In my project, I try to configure two different JPPF topology which are independent of each others in the same network. I mean, two JPPF drivers do not see each other as nodes. I examined the JPPF Manual but I didn't find any solution to solve my problem. How should I make a configuration to establish two separate networks. I will be glad if you help me to solve this problem.

Best Regards
Troubleshooting / Re: JPPF Node as a JPPF driver client
« Last post by Konstantin Eroshenko on August 09, 2017, 10:59:01 AM »

Thanks a lot. Will try it in the nearest future.
Troubleshooting / Re: JPPF Node as a JPPF driver client
« Last post by lolo on August 09, 2017, 08:41:48 AM »

Yes you can definitely create a JPPFClient and submit one or more jobs from a task executing on a node. You just need to ensure that you have enough available nodes to execute the jobs submitted from a task, otherwise the submitting task may be stuck waiting for the job to execute, and the corresponding node may remain available.

For your convenience, I attached a small sample that works like a charm for me.

Troubleshooting / Re: Task cancelling with timeout
« Last post by lolo on August 09, 2017, 06:55:29 AM »

The behavior you observe is the expected one. What happends is that your @JPPFRunnable annotated method "runTask" is performing only non-interruptible operations. When the timeout expires, the JPPF node calls Thread.interrupt() on the thread that executes the task. Since your task is not interruptible, the thread will actually never stop and will keep executing forever. As a consequence, the task never ends, and no result is ever sent back to the client, which explains why the jobReturned() and jobEnded() notfications are never sent of the client side.

To work around this, you should check the interrupted state of the executing thread, rewriting the loop like this:

Code: [Select]
public class MethodTask {
  public static String runTask(String param) {
    while (true) {
      if ("never equal".equals(param) || Thread.currentThread().isInterrupted()) {
    return "TASK FINISHED";

Note the additional test on Thread.currentThread().isInterrupted() which becomes true when the task timeout expires.

I hope this helps,
Troubleshooting / JPPF Node as a JPPF driver client
« Last post by Konstantin Eroshenko on August 08, 2017, 05:02:11 PM »
Hi, Laurent

Could you please advise the best way to utilize jppf in solving the following problem.
Some of the tasks which we send to jppf nodes takes too long time to execute. But on the node (and only on the node) we have an ability to split the task on subtasks. Unfortunately, we can't run those subtasks on one node in parallel threads. We need different processes for that. jppf nodes suits perfectly. But it's not clear, is it fine (or even possible) to make jppf node act also as jppf driver client?
Troubleshooting / Re: 1 client - 2 servers - 2 nodes topology details
« Last post by Konstantin Eroshenko on August 08, 2017, 04:47:39 PM »

I am afraid that the following is not always the case:

if driver1 and driver2 are connected to each other, then one or more PeerHandler-xxxx threads will be created in driver1 as soon as one or more nodes connect to driver2

Currently, I see in jppf console that driver1 and driver2 are connected and each has 201 nodes (one of which is a peer), but one of drivers still has no "PeerHandler-00XX" thread.
Could you, please, examine the code? Please, let me know if you won't be able to spot the issue, then I will try to find out exact scenario to reproduce it.


Sorry for false alert.
I couldn't find "PeerHandler-xxxx" threads in driver2 process in the case when driver2 is started after driver1 and all its nodes were started and connected.
driver2 will create "PeerHandler-xxxx" threads when new node connects to driver1 or already connected node become disconnected.
Also I suspected that lack of "PeerHandler-xxxx" thread prevent driver2 from knowing nodes number of driver1 and impact load balancing badly, but that's also false.
driver2 gets driver1 nodes number when they get connected. So, everything seems correct, thanks.
Pages: [1] 2 3 ... 10
JPPF Powered by SMF 2.0 RC5 | SMF © 2006–2011, Simple Machines LLC Get JPPF at Fast, secure and Free Open Source software downloads