personally id like to see more doc on monitoring big data stack ie yarn/spark -
map / reduce /executors /jobs /stages, works on port 8088 not 4040 as default would suggest
ganglia itself isnt that useful for me-i like to see the number of map tasks created/running/finished etc
some doc on optimizing map/reduce executor tasks would be nice
eg how many executors per worker node vs executor-cores given the openstack flavors in use
Chriss Gessner's comments from emails: