Scanning the tables from the shell returns this: ERROR Java::JavaIo::IOException: Unable to find region for \x03\x02\x03\x01\x01\x02\x01\x01\x01\x01\x04\x00\x04\x02\x00\x00\x02\x01\x00\x00\x01 in stones5
Scanning the meta table returns 22regions (which seems to be normal). Output is there: https://pastebin.com/axbG5C87 I will remove all HFiles and bulkload them back then I will try to reproduce the issue... Le jeu. 11 avr. 2019 à 10:44, Jean-Marc Spaggiari <jean-m...@spaggiari.org> a écrit : > Hi all, > > I don't know how i ended up in this stage, but I cleared /hbase in both ZK > and HDFS 3 days ago. created 3 brand new tables, started to put some data > on them. Twice a day I run a split command, to get a better spread. > > After 3 days I have a lot of regions overlaps. HBCK output is there: > https://pastebin.com/y4awxmN5 > > Now, how can this be repaired? If I understand correctly, -fixHdfsOverlaps > can't be used anymore. I can move the HFiles, drop the table, re-create, > reload and split again, but I have a job running, that's supposed to last 6 > days. I would have liked to keep it running. > > Any idea why there is such overlaps? > > JMS >