50 likes | 207 Views
BIGTOP. Configuration and Wrapper Executable. Motivation. Overall, even with bigtop , still lack the feel of one product family Upstream modules have inconsistent configuration mechanism Too many environment valuables to set for each modules
E N D
BIGTOP Configuration and Wrapper Executable
Motivation • Overall, even with bigtop, still lack the feel of one product family • Upstream modules have inconsistent configuration mechanism • Too many environment valuables to set for each modules • Too many commands to remember to start various module daemons • No easy way to query daemon status • All or nothing for modules – cannot add/remove modules (??? – might be useful), maybe pulling down too much or too little.
Proposal • Single BIGTOP_HOME environment to set • All other environment variables could be derived and set and listed for the user by convention. • Single executable “bigtop” for house keeping • Start/stop modules • List related settings, modules, ports, etc. • Install/remove bigtop modules
What’s Not In Scope • Wrapping upstream module commands such as hadoop, mahout, pig, etc.
Usage • export BIGTOP_HOME=/usr/lib/bigtop • /usr/bin/bigtop <options> • -h/-help – Help message, and link to bigtopGettingStarted web page • -list machines|daemons|ports|modules • -start • -stop • -add-machine <machine list> • -remove-machine <machine list> | all • -cluster sync-config|<cmd>– rsync configuration files, jar files across cluster machines • -clean – clean logs, caches across the cluster machines • -install <module> • -uninstall <module>