Thanks Jean-Daniel and Andy, Will continue reading around as much as possible - so much to understand! I thought if I wrote to a different column family, it would not affect the current scanner.
Cheers, Tim On Wed, Jan 7, 2009 at 10:35 PM, Andrew Purtell <[email protected]> wrote: > Hi Tim, > > Currently whenever you are committing back to the table in > the scanner loop you need to consider the effects of splitting. > On tap for 0.20 is a change that would allow scanning to > continue around splits in some manner (actually, the reverse) > -- but it's not yet clear how things might change client side. > Please follow HBASE-1050: > https://issues.apache.org/jira/browse/HBASE-1050 > > - Andy > >> From: Jean-Daniel Cryans >> >> I suggest you increase the hbase.regionserver.lease.period >> value from its default 1 minute to at least 2 min (120000). >> On a single node, a split can take much longer to finish >> and the client won't report in fast enough. >> >> On Wed, Jan 7, 2009 at 3:33 PM, tim robertson wrote: >> > I am seeing occasionally the following exception (not >> > reproducible reliably): >> > >> > Exception in thread "main" >> > java.lang.RuntimeException: >> > org.apache.hadoop.hbase.UnknownScannerException: > > > > >
