yes, this is an issue, I also found that. can you please file an issue?
On Sun, Sep 21, 2014 at 12:08 PM, sam liu <[email protected]> wrote: > I rollback from 2.4.1 to 2.2.0 and seems 2.2.0 does not has option > upgradeProgress, right? > > I guess it might be a hadoop issue, as I still could not start datanode > after rollback > > 2014-09-18 4:15 GMT-07:00 Susheel Kumar Gadalay <[email protected]>: > > What is the o/p of command >> >> hdfs dfsadmin -upgradeProgress status >> >> If it says upgrade is complete then you can do some sanity check by hdfs >> fsck. >> >> Stop the servers by stop-dfs.sh and then rollback by command >> start-dfs.sh -rollback >> >> On 9/18/14, sam liu <[email protected]> wrote: >> > Thanks for your comment! >> > >> > I can upgrade from 2.2.0 to 2.4.1 using command 'start-dfs.sh -upgrade', >> > however failed to rollback from 2.4.1 to 2.2.0 using command >> 'start-dfs.sh >> > -rollback': the namenode always stays on safe mode(awaiting reported >> blocks >> > (0/315)). >> > >> > Why? >> > >> > 2014-09-18 1:51 GMT-07:00 Susheel Kumar Gadalay <[email protected]>: >> > >> >> You have to upgrade both name node and data node. >> >> >> >> Better issue start-dfs.sh -upgrade. >> >> >> >> Check whether current and previous directories are present in both >> >> dfs.namenode.name.dir and dfs.datanode.data.dir directory. >> >> >> >> On 9/18/14, sam liu <[email protected]> wrote: >> >> > Hi Expert, >> >> > >> >> > Below are my steps and is it a hadoop bug or did I miss any thing? >> >> Thanks! >> >> > >> >> > Step: >> >> > [A] Upgrade >> >> > 1. Install Hadoop 2.2.0 cluster >> >> > 2. Stop Hadoop services >> >> > 3. Replace 2.2.0 binaries with 2.4.1 binaries >> >> > 4. Start datanodes: $HADOOP_HOME/sbin/hadoop-daemon.sh start datanode >> >> > 5. Start namenode with option upgrade: >> >> > $HADOOP_HOME/sbin/hadoop-daemon.sh >> >> > start namenode -upgrade >> >> > 6. Start secondary namenode, tasktracker and jobtracker >> >> > >> >> > Result: >> >> > >> >> > Whole upgrade process could be completed successfully. >> >> > >> >> > [B] Rollback >> >> > 1. Stop all hadoop services >> >> > 2. Replace 2.4.1 binaries with 2.2.0 binaries >> >> > 3. Start datanodes: $HADOOP_HOME/sbin/hadoop-daemon.sh start datanode >> >> > 4. Start namenode with option upgrade: >> >> > $HADOOP_HOME/sbin/hadoop-daemon.sh >> >> > start namenode -rollback >> >> > >> >> > Result: >> >> > >> >> > Namenode service could be started >> >> > Datanodes failed with exception: >> >> > Issue: DataNode failed with following exception >> >> > 2014-09-17 11:04:51,416 INFO >> >> > org.apache.hadoop.hdfs.server.common.Storage: Lock on >> >> > /hadoop/hdfs/data/in_use.lock acquired by nodename >> >> > 817443@shihc071-public >> >> > 2014-09-17 11:04:51,418 FATAL >> >> > org.apache.hadoop.hdfs.server.datanode.DataNode: Initialization >> failed >> >> for >> >> > block pool Block pool BP-977402492-9.181.64.185-1410497086460 >> (storage >> >> id ) >> >> > service to hostname/ip:9000 >> >> > org.apache.hadoop.hdfs.server.common.IncorrectVersionException: >> >> > Unexpected version of storage directory /hadoop/hdfs/data. Reported: >> >> > -55. >> >> > Expecting = -47. >> >> > at >> >> > >> >> >> org.apache.hadoop.hdfs.server.common.Storage.setLayoutVersion(Storage.java:1082) >> >> > at >> >> > >> >> >> org.apache.hadoop.hdfs.server.datanode.DataStorage.setFieldsFromProperties(DataStorage.java:302) >> >> > at >> >> > >> >> >> org.apache.hadoop.hdfs.server.common.Storage.readProperties(Storage.java:921) >> >> > >> >> >> > >> > >
