Install jblas linear algebra lib in MLlib

14/12/20 20:33:52 WARN TaskSetManager: Loss was due to java.lang.UnsatisfiedLinkError
java.lang.UnsatisfiedLinkError: org.jblas.NativeBlas.dposv(CII[DII[DII)I

MLLib uses native libraries, which need to be present on the nodes. (that is it does not come with spark installation)

MLlib uses the jblas linear algebra library, which itself depends on native Fortran routines. You may need to install the gfortran runtime library if it is not already present on your nodes. MLlib will throw a linking error if it cannot detect these libraries automatically.To use MLlib in Python, you will need NumPy version 1.7 or newer and Python 2.7.

You have to make sure that libgfortran library exists on all nodes.

for debian/ubuntu use: sudo apt-get install libgfortran3

for centos/redhat use: sudo yum install gcc-gfortran



Reference:

http://stackoverflow.com/questions/24758314/apache-spark-mllib-collaborative-filtering