FWIW: I would never directly delete files from the hbase.rootdir. Past experience has told me that if files are not being cleaned up, then there is either a bug, or a valid reason for keeping the files. I would dig into why the logcleaner is not removing the files. Could be things like replication between two clusters not working. I believe the default (depending on version) is to have the master logcleaner wake up every 10 minutes. You should be able to set debug logging, and figure out why the logs are not being expunged.
On Mon, Jul 14, 2014 at 10:03 AM, Ted Yu <[email protected]> wrote: > What are the values for the following config in your cluster ? > hbase.master.logcleaner.plugins > hbase.master.logcleaner.ttl > > BTW in 0.98, -ROOT- is gone. > > Cheers > > > On Mon, Jul 14, 2014 at 9:43 AM, Giuseppe Reina <[email protected]> wrote: > > > Hi all, > > after running a write intensive workload on HBase (0.98) I exhausted > the > > HDFS space. I realized that most of the HDFS space was occupied by files > in > > the path /apps/hbase/data/oldWALs/ > > Apparently either the old log files were not cleaned quickly enough or > they > > were not cleaned at all. I'm wondering if I can safely delete those > files > > and if I can how can I do it? Manually or is there an "hbase" command? > > > > > > > > Also as an unrelated question, where are stored the -ROOT- and -META- > > information in zookeeper? > > > > > > Kind Regards > > >
