[ http://issues.apache.org/jira/browse/HADOOP-514?page=comments#action_12433496 ] Yoram Arnon commented on HADOOP-514: ------------------------------------
agreed, the dfs needs to be responsive. but the larger the system, the more frequently datanodes hit the name node. a block that awaits re-replication will get replicated to the first node that sends a heartbeat and makes sense as a replication destination. Overall the system is very responsive even though each individual data node sends infrequent heartbeats. as for deletions, they can wait a bit, especially with undelete functionality in mind. The formula max(nodes/10, 1) maintains heartbeat rate at a constant 10 per second at the namenode. We can tweak the denominator to change that rate, but this seems like a good rate to get responsiveness, while limiting the load on the namenode. > namenode heartbeat interval should be configurable > -------------------------------------------------- > > Key: HADOOP-514 > URL: http://issues.apache.org/jira/browse/HADOOP-514 > Project: Hadoop > Issue Type: Bug > Components: dfs > Reporter: Wendy Chien > > The namenode heartbeat interval is currently a constant in FSConstants. It > should be configurable instead. Things which currently rely on the heartbeat > occuring fairly frequently should be disassociated. -- 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
