Hi, There is 2 reasons to have a major compaction.
the first on is when a minor compaction selects all the files to be compacted. It is then promoted as a major compaction for that region. The second reason is time based. Every day or week, depending on your configuration, HBase will trigger a major compaction. In your case, what it can be: - Flushs are coming too fast and are too small and compaction can not keep up. - Major compaction has been disabled in the configuration and is never triggered manually. - Major compaction are spiked for another reason that can be seen in the logs. Are you able to share the logs for one region server? What happens when you try to trigger a major compaction from the shell? JMS 2016-08-17 5:14 GMT-04:00 吴国泉wgq <wgq...@qunar.com>: > Hi ALL: > > Today I met a problem, There is a very large table in the hbase > cluster, And this table never ran a major-compaction and I trigger a > major-compaction. > > The major-compaction will take a very very long time, meantime, > this table is still being writing. > > The number of the Hfile is increasing, After a period of time, the > writing is blocked because some region have too many Hfiles. > > I wonder If hbase can trigger minor-compaction when it is doing > major-compaction? > > If cannot, How can I solve this problem. > >
