一点收获：
hadoop streaming:
ignore map failure: -D mapred.max.map.failures.percent=10 allows 10 percent of map failure
bzip2 check:
bzip2 -t filename
shell extended glob:
shopt -u extglob
shopt -s extglob
hdfs glob:
from api: org.apache.hadoop.fs.GlobFilter A filter for POSIX glob pattern with brace expansions
