Did you get the name of the broken reference? I'd trace its life in namenode logs and in regionserver log by searching its name (You might have to find the region in master logs to see where region landed over time). The reference name includes the encoded region name as a suffix. This is the region that the reference 'references'.... so need to figure what happened with it. Did it get cleaned up before reference was cleared? (Something that should not happen).
St.Ack On Mon, Jan 21, 2013 at 2:20 PM, Jean-Marc Spaggiari < jean-m...@spaggiari.org> wrote: > Hum. It's still a bit obscur for me how this happend to my cluster... > > -repair helped to fix that, so I'm now fine. I will re-run the job I > ran and see if this is happening again. > > Thanks, > > JM > > 2013/1/21, Stack <st...@duboce.net>: > > On Mon, Jan 21, 2013 at 12:01 PM, Jean-Marc Spaggiari < > > jean-m...@spaggiari.org> wrote: > > > >> Found lingering reference file > > > > > > > > The comment on the method that is finding the lingering reference files > is > > pretty good: > > > http://hbase.apache.org/xref/org/apache/hadoop/hbase/util/HBaseFsck.html#604 > > > > It looks like a reference file that lost its referencee. > > > > If you pass this arg., does it help? > > > > > http://hbase.apache.org/xref/org/apache/hadoop/hbase/util/HBaseFsck.html#3391 > > > > > > St.Ack > > >
