To Build

$ mvn clean install

To Run with Docker

$ docker run --rm -p 8001:8181 opennms/distributed-scheduling-poc:local 

# Run with a console and attach to it
$ docker run -it -a STDIN -a STDOUT -a STDERR --rm -p 8001:8181 opennms/distributed-scheduling-poc:local --

# Run with debugging enabled
$ docker run -it -a STDIN -a STDOUT -a STDERR --rm -p 5005:5005 -p 8001:8181 opennms/distributed-scheduling-poc:local debug

To Run in Karaf Native

Karaf Native = running Karaf directly from the O/S, instead of containerized or within a VM.

First you must patch your karaf bash script to account some odd reflection that Ignite is doing. Look for this section:

  if [ "${VERSION}" -gt "8" ]; then

And then add this in there

  --add-opens java.base/java.nio=ALL-UNNAMED \

Note that without this setting, Ignite fails fast with an error such as:

  Caused by: java.lang.reflect.InaccessibleObjectException: Unable to make field long java.nio.Buffer.address accessible: module java.base does not "opens java.nio" to unnamed module @31ce8fe7

Also note this problem does not appear to occur on JDK 11, but it does on 17 and 18.

And then run these commands

  ./bin/karaf
  feature:repo-add mvn:org.opennms.poc/features/1.0.0-SNAPSHOT/xml/features
  feature:install ignite-poc

To run the Detector and Monitor plugins

  feature:install icmp-plugins
  feature:install snmp-plugins

Using the example docker-compose file

WARNING: the docker-compose file is out-of-date

$ cd tools
$ docker-compose up

Using Skaffold

Requires Skaffold and Kind (or other Kubernetes server)

$ skaffold dev

To Exercise

Wait for Twin GRPC startup. The following log message indicates Twin GRPC has started:

[poc-ignite-worker] 00:46:34.914 INFO  [Blueprint Extender: 2] Started Twin gRPC Subscriber at location cloud with systemId 0ddba11

Here is a 1-liner to watch for this line (note this command only shows logs from a single pod):

watch -c 'kubectl logs --all-containers=true deployment/poc-distributed-scheduling | grep "Started Twin gRPC"'

Send a workflow update via the test-driver:

$ cd tools
$ ./post-workflow

Play with workflows

$ curl 'http://localhost:8001/poc/load-em-up?size=SMALL'

UNSOLVED

Adding linkerd containers to the ignite application pods interferes with the TcpDiscoveryKubernetesIpFinder.
- Nodes fail to cluster together
- See log errors such as:
  
  Caused by: class org.apache.ignite.spi.IgniteSpiOperationTimeoutException: Failed to perform handshake due to timeout (consider increasing 'connectionTimeout' configuration property).

opennms-forge/distributed-scheduling-poc