Hi David, I agree, that should be checked if even just for the sake of completeness. Can you please open a jira?
Thanks, J-D On Wed, Jul 27, 2011 at 6:02 PM, David Capwell <[email protected]> wrote: > Heya, I am testing hbase with bulk loads and I seeing something unexpected. > > I'm generating a set of random KeyValues where key, family, column, and value > are all random strings, i then sort them as Arrays.sort(values, > KeyValue.COMPARATOR); > I wrote this list to a StoreFile.Writer under /tmp/$tableName/$family and > tried to load it into HBase > > LoadIncrementalHFiles loadIncrementalHFiles = new > LoadIncrementalHFiles(this.conf); > loadIncrementalHFiles.doBulkLoad(path, table); // path is /tmp/$tableName > > The store file was not rejected and when I scan the table later with just the > family, it seems to return all the randomly generated values (even though > they don't belong to this family) > Scan scan = new Scan(); > scan.addFamily(this.family); > ResultScanner scanner = table.getScanner(scan); > for(Result r : scanner) { > for(KeyValue kv: r.list()) { > KeyValueUtil.print(kv); > } > } > > Based off http://hbase.apache.org/bulk-loads.html LoadIncrementalHFile will > determine which region the HFile belongs to; should it also check that the > hfile belongs to the right family? > > I am running hbase-0.90.3 on 10 nodes and hadoop-0.20.204.1 on the same 10 > nodes > > Thanks for your time reading this email.
