Maybe stack will chime in here with a potential recovery mechanism, but Erik is correct. What has happened is the metadata which indicates which tables exist, and what their ranges are has disappeared. Right now there is no easy way to recover back to the original because of the missing metadata is not stored elsewhere. What you can try in the mean time is directly accessing the mapfiles using the raw mapfile reader (i think the class is Mapfile...) - you'd be able to get the data out, then you could re-insert it back into a running instance later.
Needless to say, one should not trust a 1-node cluster to irreplacable data. Until certain HDFS bugs are resolved, which are slated for a hbase 0.20 timeline, there is always a data loss hole. Good news is there might be a backported HDFS 0.19 patch, but that may not be relevant since HBase 0.20 is based on Hadoop/HDFS 0.20. good luck... -ryan On Mon, May 25, 2009 at 12:24 PM, Erik Holstad <[email protected]>wrote: > Hey Arber! > What it sounds like to me is that the table Meta hadn't been flushed to > disk > and was inly sitting on memory, so > when the machine went down that data got lost. > > Regards Erik >
