[ http://issues.apache.org/jira/browse/HADOOP-94?page=comments#action_12377277 ]
Doug Cutting commented on HADOOP-94: ------------------------------------ I agree that we should put some sort of a lock file in the data directory, but I don't think we should move the existing pid file, since that is managed by the generic daemon start/stop code. Rather we could use the nio file locking code to create an exclusive lock on a file. This will automatically be unlocked by the kernel if/when the jvm exits. > disallow more than one datanode running on one computing sharing the same > data directory > ---------------------------------------------------------------------------------------- > > Key: HADOOP-94 > URL: http://issues.apache.org/jira/browse/HADOOP-94 > Project: Hadoop > Type: Bug > Components: dfs > Versions: 0.2 > Reporter: Hairong Kuang > Fix For: 0.3 > > Currently dfs disallows more one datanode to run on the same computer if they > are started up using the same hadoop conf dir. However, this does not prevent > more than one data node gets started, each using a different conf dir > (strickly speaking, a different pid file). If every machine has two such > datanodes running, namenode will be busy on deleting and replicating blocks > or eventually lead to block loss. > Suggested solution: put pid file in the data directory and disallow > configuration. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://issues.apache.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira
