I'm glad to follow up to my post telling you that regions which are down go up to the other nodes after about 20 minutes! I've only have to wait a little more than i did...
But there is no a way to reduce this time? I mean, about 20 minutes to wait the regions recovered is too long! There is not a property for hdfs-site or hbase-site in which I could tell set this time? Thank for the help!