hadoop - Block assignation using network topology -
if understood principles when applying network topology, blocks written:
- on client server if hosting datanode
- on second server defined on different rack
- on third server defined on same rack #2
is policy configurable or “hard-written” in class? of course, not want modify class myself…
basically, to:
take account datacenter (according read, hdfs not care datacenters if using network topology)
force write in 3 distinct racks
how do that?
there capability override baseline block allocation algorithm involve writing quite bit of java code , there aren't real examples out there. here blog link jira ticket explaining enhancement:
Comments
Post a Comment