There are certain instances in which a task must be run in the background of BigConnect. Some typical reasons to run a background task in BigConnect are:
Incoming data needs to be run outside of the data worker framework,
Work done by a web request would take too long and would ruin the UX,
You would like to spawn an asynchronous task a long-running task inside of BigConnect that will notify the web UI when it is done but not block on a request.
BigConnect supports Long Running Processes in the background. Long Running Processes are similar to Data Workers: they recieve a message from a queue, the Long Running Process decides if it can operate on the message, and then works on that message. An example of a Long Running Process is FindPathLongRunningProcessWorker. When a find-path request is issued from the graph on the UI, a web request is made to the server with the source and destination vertex and a message is queued into the configured long running process queue. That message is then dequeued by all of the current long running processes and the FindPathLongRunningProcessWorker sees that it can work on the message, finds all of the paths between the specified vertices, and sends those results to the UI.
Long Running Processes have an
isHandled method and a
processInternal method that need to be implemented as a minimum to create a LongRunningProcessWorker. Like the Data Workers, the
isHandled method allows the long running process to filter through the messages that are put onto the long running process queue. Then, the canonical class reference must be in that same project's
META-INF/services/com.mware.core.model.longRunningProcess.LongRunningProcessWorker file so it can be loaded from the classpath.
This method returns true if the Long Running Process Worker can operate on the incoming message. Otherwise, it returns false.
This method handles the message that the isHandled method above returned true for.
Long Running Processes are deployed inside of the web server by default. Running the Long Running Processes in the web server is good enough for a typical installation, but the requirements are going to vary widely depending on things like:
Your data size,
How resource-intensive your long running processes are on the server,
The number of cores on each box.
Long Running Processes are run inside of their own threads on the server and run independently of each other. It is possible to run more than one instance of long running processes which can be useful if your Long Running Process message queue gets backed up.
To add more instances of each Long Running Process Worker, simply add/change the following in your properties file:
#Set number of long running process threads to 2com.mware.core.process.LongRunningProcessRunnerProcess.threadCount=2
If you do not want Long Running Processes to run inside of the web server, add the following to your configuration:
#disable the long running processes running inside of the web serverdisable.com.mware.core.process.LongRunningProcessRunnerProcess=true