radamsa-haproxy

This provides a Dockerfile for load balancing multiple Radamsa servers under HAProxy.

In situations where Radamsa is being used as a server and is a CPU bottleneck, this might be useful. Multiple independent servers are run, so it only makes sense when using Radamsa for non-deterministic fuzzing.

This was a quick solution for a very specific problem I encountered on a fuzzing project. While not as elegant, it was a much simpler solution than modifying Radamsa itself. Since HAProxy provides a single port, applications that use a Radamsa server don't need any configuration change to use this instead.

Usage Examples:

# Build a Docker image:
docker build -t radamsa-haproxy .

# Run on port 8888 with the directory "my-samples" containing input files
docker run \
    --rm \
    --init \
    -v `pwd`/my-samples:/samples \
    -p 8888:8888 \
    radamsa-haproxy

A directory must be mounted into /samples containing one or more files to use as input to the fuzzer. In the example, pwd is used to insert the current working directory, since Docker requires absolute paths for mounts.

By default 4 Radamsa servers are started. The number can be changed by passing NUM_SERVERS into the container.