Namenode does not process block report if the rack-location script is not
provided on namenode
----------------------------------------------------------------------------------------------
Key: HADOOP-3065
URL: https://issues.apache.org/jira/browse/HADOOP-3065
Project: Hadoop Core
Issue Type: Bug
Components: dfs
Reporter: dhruba borthakur
Assignee: Devaraj Das
Per HADOOP-1985, If the namenode does not have a dnsToSwicthMapping specified,
it does not process block reports from datanode(s). In the ResolutionMonitor,
we have the following piece of code:
List<String> rName = dnsToSwitchMapping.resolve(dnHosts);
if (rName == null) {
continue;
}
Instead, we should probably set "DEFAULT RACK" for this datanode.
Also, when a processBlockRepont command is dropped by the namenode, it should
log a message to facilitate debugging. The code in FSNamesystem.processReport
could be something like this:
if (node.getNetworkLocation().equals(NetworkTopology.UNRESOLVED)) {
LOG.info("Ignoring block report from " + nodeID.getName() +
" because rack location for this datanode is still to be
resolved.");
return null; //drop the block report if the dn hasn't been resolved
}
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
