Public Repository

Last pushed: a year ago
Short Description
Generating stats about public model repositories using the BiVeS tool.
Full Description

The statsgenerator is a Java tool, stored in /statsgenerator.jar of that image.
By default, the statsgenerator will write the data into the following locations inside the container:

  • crawled models /srv/modelstats/storage/${REPOSITORY}
  • generated statistics to /srv/modelstats/storage/stats
  • temporary/working files to /srv/modelstats/working

Thus, you may want to mount some persitent space at least to /srv/modelstats/storage.

A typical run will therefore look like:

docker run -it --rm -v /persistent/space:/srv/modelstats/ binfalse/bives-statsgenerator:latest java -jar /statsgenerator.jar

To get more information you may add the -h flag to the java call.

Issues

As the repositories are quite big and models in it tend to become more complex you may run into Java memory issues. In that case you could pass the following flags to the Java runtime environment:

  • -XX:-UseGCOverheadLimit to remove the garbage collectors overhead limit -- sometimes the GC has problems removing all the tiny things and runs into an "overhead" because he cannot free memory fast enough
  • -XX:+UseParallelGC do garbage collection in parallel
  • -Xms4096M give the JVM 4GB of mem initially
  • -Xmx65536M allow up to 64GB of mem

For example:

docker run -it --rm -v /persistent/space:/srv/modelstats/ binfalse/bives-statsgenerator:latest java -XX:-UseGCOverheadLimit -XX:+UseParallelGC -Xms65536M -Xmx65536M -jar /statsgenerator.jar
Docker Pull Command
Owner
binfalse