Configuration of triggers in RapidMiner Server

RichL

Hello everybody,

while getting familiar with RM Server I´m now trying to automate some processes through triggers, with either "Monitor File Changes" or "Incoming Emails".

However I don´t find any documentation on how to properly configure it. For example, when monitoring file changes, the directory of the file is needed. when i paste it in, it only says "monitoring unknown path". So what am I doing wrong here...?

At the E-Mail config, a JNDI reference is needed. My research didn´t clarify that either, so that´s why i´d ask for help here!

Thank you in advance!

Richard

mmichel

Hi RichL,

I managed to gather some additional knowledge, renamed the JNDI in the config-file of the server, but unfortunately there´s a NullPointerException that i get when I try to toggle the trigger.

It seems that the JNDI lookup is not working correctly. I will create an internal issue and we'll have a look on it. Thanks for the feedback.

Unfortunately though, even with a process like that, the triggering is not working.

File triggers are only supposed to work directly on the file system. So, you need to refer to a real path (not a repository path). Something like /tmp/data/csv (in case of unix os).

Cheers,
Marcel

sgenzer

cc @mmichel

RichL

Hey Scott!

Thanks for the reply!

I managed to gather some additional knowledge, renamed the JNDI in the config-file of the server, but unfortunately there´s a NullPointerException that i get when I try to toggle the trigger.

I already talked a bit about this topic with a colleague of yours, but additional input is always welcome!

Best regards,

Richard

David_A

Hi @RichL ,

I have just recently created to sample processes for using the "file change" triggers on server.

You have basically two options: You can monitor a single file for changes (for example, another program writes into that file, or a log is dumped always dumped at the same place with the same name).

In this case you can simply create a process that reads said file and processes it further.

The second case is you monitor a complete folder for new files or changes. In this case you need a more generic process that uses the special macro %{changed_file_path} which contains the complete path to the file that was changed. In case of several changed files, the process is triggered for each file separately (unless you have selected "Batch mode" when creating the trigger, see below). The batch mode is a bit more tricky, as all changed files are collected and the paths are stored in the  macro as a newline separated list of all changed file paths.

In general I would only recommend the triggers for events that don't happen too frequent, as the monitoring is of course not 100% real time.

Unfortunately I can't help you with the mail trigger, as I haven't used those myself.

As a general hint, the activation of the triggers is a bit hidden. One need to click the red/green bubble in the interface to toggle the status.

Best,
David

RichL

Hey @David_A ,

many thanks for the reply.

So you would say, that I should schedule the process, that saves the csv to the server, monitor that file or directory and build a trigger to excecute it, as soon as anything changes, right? Because I cannot monitor anything away from the server, as far as I know at least...

Unfortunately though, even with a process like that, the triggering is not working.

Best regards,

Richard

RichL

Hey Marcel!

Wow, thanks! It finally worked I have a local server up and running right now, so I placed the csv into the server directory on my harddrive and copied the link to the server´s interface. I tested, it also worked with different file directories on my local machine.

My question now would be, how to make this work when the file is stored on an actual server? Can i use relative paths as well to point to the directory to be monitored?

Best

Richard

mmichel

Hi RichL,

glad that it is working now

My question now would be, how to make this work when the file is stored on an actual server? Can i use relative paths as well to point to the directory to be monitored?

You either need to use the same path or you need to adapt the triggers per server instance. Relative paths are also working, but these depend on the execution directory. So, if you are starting server with just ./standalone.sh your base directory will look like this $installDir/bin, and if you are starting server via ./bin/standalone.sh your base directory will be $installDir. So, I would recommend absolute paths here, as relative paths might change easily.

Cheers,
Marcel