We have upgraded to the phoenix on master. What we did is to drop system.table, install new jar, restart regionservers, rerun the ddl.
It seems to go wtihout a hitch. But we noticed that the write to phoenix table is getting extremely slow. It seems that the regionservers are busying archiving the old table. Here is a snippet of hbase log: 014-02-27 20:42:03,553 DEBUG org.apache.hadoop.hbase.backup.HFileArchiver: Finished archiving file from: class org.apache.hadoop.hbase.backup.HFileArchiver$FileableStoreFile, file:hdfs://bicluster/hbase/BI.EVENTS/f25c0f2022a25323043933ba61b8cd92/USER/a50f3522f5f044448509a6cc143c4ae9, to: hdfs://bicluster/hbase/.archive/BI.EVENTS/f25c0f2022a25323043933ba61b8cd92/USER/a50f3522f5f044448509a6cc143c4ae9 2014-02-27 20:42:03,553 DEBUG org.apache.hadoop.hbase.backup.HFileArchiver: Archiving:class org.apache.hadoop.hbase.backup.HFileArchiver$FileableStoreFile, file:hdfs://bicluster/hbase/BI.EVENTS/f25c0f2022a25323043933ba61b8cd92/USER/d184255d968a4885895af6504162dae6 Also seeing a whole bunch of 014-02-27 20:53:50,766 DEBUG org.apache.hadoop.hbase.index.IndexLogRollSynchronizer: Releasing INDEX_UPDATE writelock 2014-02-27 20:53:50,766 DEBUG org.apache.hadoop.hbase.index.IndexLogRollSynchronizer: Releasing INDEX_UPDATE writelock Just wondering what exactly is going on here? Is this the expected behavior? And how long this whole process will take for the system to be back to normal? Thanks Sean
