Print

[benlawry]

Sorry I've had SOOOOOO many questions lately - I'm hoping this is the last one, relating to JPPF 1.9 configuration, anyway

So we are pretty much finished with the upgrade from 1.6 to 1.9 - I can run some example programs and the tasks seem to execute fine.  My current problem is, I'm having an issue with the JPPF Admin GUI (almost certainly a config file issue - I must be shooting myself in the foot somehow).  What seems to happen is, the Admin GUI can connect to the driver in a degraded capacity - it can tell that we have 144 nodes running, but it doesn't display them.  IE, in the list of connected nodes (where it usually shows a hostname/IP address and a port number), it only shows two connected nodes and the status for each is unknown.

I'll attach a copy of my config files below - the only difference between these and what we were running before is we have jppf.discovery.enabled set in these.  I've tried different combinations of setting these to true or false.  I can't seem to run nodes if autodiscovery is enabled on them, so I've been enabling autodiscovery only on the client/admin and driver:

Driver config:

app.server.port = 14000
class.server.port = 14100
jppf.discovery.enabled = true
jppf.management.port = 14300
jppf.management.rmi.port = 14400
jppf.object.input.stream.class = java.io.ObjectInputStream
jppf.object.output.stream.class = java.io.ObjectOutputStream
jppf.object.stream.builder = org.jppf.serialization.XstreamObjectStreamBuilder
jppf.peer.discovery.enabled = false
max.memory.option = 768
node.server.port = 14200
remote.debug.enabled = true
remote.debug.port = 14500
strategy.test.decreaseRatio = 0.2
strategy.test.discountFactor = 0.2
strategy.test.increaseRate = 0.03
strategy.test.maxDeviation = 0.2
strategy.test.maxGuessToStable = 50
strategy.test.minSamplesToAnalyse = 100
strategy.test.minSamplesToCheckConvergence = 50
strategy.test.performanceCacheSize = 2000
strategy.test.proportionalityFactor = 2
strategy.test.rateOfChange = 0.9
strategy.test.sizeRatioDeviation = 1.5
task.bundle.autotuned.strategy = test
task.bundle.size = 1
task.bundle.strategy = proportional
transition.thread.pool.size = 1

Node Config:

app.server.port = 14000
class.server.port = 14100
jppf.discovery.enabled = false
jppf.management.port = 16000
jppf.management.rmi.port = 17000
jppf.server.host = newton1
max.memory.option = 128
node.server.port = 14200
processing.threads = 1
reconnect.max.time = 5
remote.debug.enabled = false
remote.debug.port = 15000

Admin Config:

jppf.discovery.enabled = true
jppf.drivers = AwesomeDriver
AwesomeDriver.jppf.server.host = 134.253.123.5
AwesomeDriver.class.server.port = 14100
AwesomeDriver.app.server.port = 14000
AwesomeDriver.jppf.management.port = 14300
AwesomeDriver.jppf.management.rmi.port = 14400
AwesomeDriver.jppf.discovery.enabled = true
AwesomeDriver.priority = 10
AwesomeDriver.jppf.management.host = 134.253.123.5
reconnect.max.time = 3600
default.refresh.interval = 1000

[lolo]

Hello Ben,

Sorry I've had SOOOOOO many questions lately

Please don't be sorry, I am really grateful for your questions, as they contribute greatly to improving JPPF.

To help understand the situation, I would like to gather some information on which host address the nodes are using for the management features.
Could you please take one of the nodes you can't see in the console and modify its logging configuration as follows (I'm assuming you're using Log4j):
  add this line at the end of the log4j-node.properties file: "log4j.logger.org.jppf.utils.NetworkUtils=DEBUG"
Then start the node and look into its log file, you should see at least 2 messages as follows:

Code: [Select]

JMX host from NetworkUtils: <some ip address>
computed JMX host: <some ip address>

Could you please let us know if the IP address in the log is reachable from the admin console host?
Can you also let us know if your nodes hosts have a single or multiple network interfaces, whether physical or virtual (such as a VPN connection)?

I have also noticed, from the configurations you posted, that you are using "134.253.123.5" on the client/GUI side, and "newton1" on the node side. Is "134.253.123.5" the IP address "newton1" resolves to on the client side, and vice-versa?

Also, if you want to enable auto-discovery on the nodes with capability to fall back to the values configured in the properties file, you can do the following:
- enable auto-discovery "jppf.discovery.enabled = true"
- set the property "jppf.discovery.timeout" to a low value, e.g. a few seconds. The default value is 5000 milliseconds. This is the time after which the node will stop using auto-discovery and fall back to the configuration file.

Thank you very much for your time,
-Laurent

[benlawry]

Thanks for the suggestions - unfortunately, we've had some setbacks in our own codebase that I need to take care of for now - I'll try your suggestions as soon as I can and reply as soon as I get a chance.