I am glad that I could help.

In our case, we followed mostly the configuration from here:
it a bit to adapt to our requirements e.g. today we run 2GB containers
instead of 3-4GB, but it might change in the future). Make also sure that
memorry allocated in mapreduce.map.java.opts is smaller than
mapreduce.map.memory.mb (the same for reduce tasks).
2013/12/11 Silvina Caíno Lores <[EMAIL PROTECTED]>