Hi, Is it possible to use IP addresses for the host list of BlockLocation? The Ceph file system uses IPs to describe block locations, and Hadoop is not achieving any data locality. I used a quick reverse DNS hack to verify that the IPs were causing the problem, but this isn't robust, and the Ceph file system cannot report host names.
It looks like the issue crops up in comparing Nodes in JobInProgress:getMatchingLevelForNodes where one node is /default-rack/DNS-NAME and the other is /default-rack/IP-address, and equality check fails. Is there a configuration step I am missing, or is this simply not supported? Are there any suggestions on how to handle this? Thanks! Noah