amazon web services - It is possible use AutoScaling with Elastic Mapreduce? -
i know if can use autoscaling automatically scaling or down amazon ec2 capacity according cpu utilization elastic map reduce.
for example, start mapreduce job 1 instance, if instance arrive 50% utilization example want use created autoscaling group start new instance. possible?
do know if possible? or elastic mapreduce because "elastic", if needs starts automatically more instances without configuration?
no, auto scaling cannot used amazon elastic mapreduce (emr).
it possible scale emr via api or command-line calls, adding , removing task nodes (which not host hdfs storage). note not possible remove core nodes (because host hdfs storage, , removing nodes lead lost data). in fact, difference between core , task nodes.
it possible change number of nodes within emr "step". steps executed sequentially, cluster made larger prior step requiring heavy processing, , reduced in size in subsequent step.
from emr developer guide:
you can have different number of slave nodes each cluster step. can add step running cluster modify number of slave nodes. because steps guaranteed run sequentially default, can specify number of running slave nodes step.
cpu not metric on base scaling of emr cluster, since hadoop keep nodes busy possible when job running. better metric number of jobs waiting, can finish quicker.
see also:
Comments
Post a Comment