Having config file to map out the output filename template is nice. However, I think it is probably secondary. I think having parser decide where to put the filename is a good start.
Regards, Eric On 3/10/10 10:02 AM, "Guillermo Pérez" <bi...@tuenti.com> wrote: > Right now the conf/chukwa-demux-conf.xml file defines a set of aliases > for mappers, so you just specify a Name of the record, and the > appropriate mapper class is used. I would like to do more or less the > same for reducers. > > My mapper generates different chukwa records with different keys. For > example record "network" and record "disk". But the reducer class is > the same, just makes some calculations. Right now if I specify the > same reducer class in the mapper, the output file is the name of the > reducer, so both types of records end in the same file. > > I would like to just make key.setReduceType(ReduceAlias). That > ReduceAlias will be the name of the output file, but the actual class > for the reduce will be defined in the config, the same way we do for > mappers. Or maybe -again- setting the output file name in the key as I > suggested for the cluster destination, but I think the first will be > more general, but may tie code to config (a mapper sets reduce type > "foo" in code, and needs "foo" to be defined in the config) . What do > you think? What will be the best approach?
