FPGrowth doesnt use Hadoop. Its in-memory implementation of the algo. You need to increase the java heapspace of your client. (if the data can fit the memory)
Robin On Wed, May 25, 2011 at 1:44 PM, Sean Owen <[email protected]> wrote: > I don't think setting this on the command line does anything. It affects > the > client JVM not the cluster. Set this in find/makers-site.conf > On May 25, 2011 2:42 AM, "Vipul Pandey" <[email protected]> wrote: > > hi! > > > > I had been running PFPGrowth embedded in my code so far in that my main > > class launches PFPGrowth after accepting the config params from command > > line. I used to set -Dmapred.child.java.opts="-server -Xmx2G -Xms2G" for > my > > jobs and the config used to be passed along to PFPGrowth as well. > > I just switched to using FPGrowthDriver directly to launch my jobs and > now > > my tasks are failing due to OOME. It doesn't allow me to set the options > as > > above. > > Any way out? > > > > thanks! > > vipul >
