Ubuntu 18.04 RapidMiner 9.6 runs out of sockets

jacobcybulski

For a while I had an issue that after some hours or days of use my system would stop opening windows. I assume it was running out of memory, even thought it has 64Gb RAM. Then I observed that this happens when I use RM extensively. After some investigation, I found that RM creates an executable program "browsercore64" inside the folder ".RapidMiner/Internal cache/browser7.4-ANY", and while RM stay alive "browsercore64" creates files or sockets until things go bad. When I run "lsof" I can see that the number of files associated with "browsercore64" rarely goes down, with every run and every visualisation, it creeps up into several hundreds. Then I cannot open any more windows even to investigate what is going on. I have never experienced this problem with the previous versions of RM (or may be I wasn't using it as heavily as recently). In Win 10 world some people consider "browsercore64" evil and a trojan, however, I suspect it is part of chromium library allowing RM to display results.

At the moment, every so often I have to shut down RM which frees the sockets. Is this a problem with my system, Java, RM setup or perhaps a bug in RapidMiner 9.6?

Jacob

Marco_Boeck

Hi,

The code closes each browser instance when a result tab is closed. I do not know why the library fails to free the resources properly on Linux.
We will update the browser library with version 9.7, hopefully that is going to fix the problem. If not, please let us know, so we can create a bug report for the libary..

Regards,
Marco

sgenzer

Hi Jacob - hmm that's definitely a question for@Marco_Boeck. It's a holiday in Europe today (May 1) but hopefully he can share some thoughts next week.

Scott

jacobcybulski

Thanks@sgenzer,这就太好了e issues is very persistent and in my case repeatable. When 250 of "browsercore64" are listed by lsof, either RM becomes unresponsive or new windows do not pop up. Just in case, I attach a fragment of my system log around the time it happens.