Permalink
Find file
Fetching contributors…
Cannot retrieve contributors at this time
59 lines (37 sloc) 1.18 KB
Reimplementing some of the hadoop-ec2 tools without so much shell script as a way to get cozy with the AWS SDK.
Building -
it's currently setup as a Netbeans project, will be getting the Netbeans stuff out soon.
depends on:
--aws-sdk
http://aws.amazon.com/sdkforjava/
--commons-codec v. 1.4 (aws includes 1.3 but I had issues using it)
http://commons.apache.org/codec/
--log4j
http://logging.apache.org/log4j/
--ganymed (ssh lib)
http://www.cleondris.ch/opensource/ssh2/
setup the required libraries and build.
Configuring --
there's sample config files in dist/config
to launch a cluster you need to choose or build an AMI, preferably with hadoop installed, but you could install hadoop on startup without too much effort.
Running --
run pooper.sh in dist/bin after building.
As of this writing, HaddooperPooper 's command-list:
command
create-groups
delete-cluster
describe-cluster
get-login-command
launch-cluster
launch-master
launch-slaves
list-clusters
login
push-file
terminate-cluster
terminate-slaves
Still to come:
port forwarding,
automated job submission/termination,
support for a map-reduce only (non-dfs) slave group that can be scaled separately,
GUI.