We are using Hunk with MapR. There is a dispatch directory that Hunk uses for the reduce of the map reduce. /mapr/tmp/dispatch
.
We are seeing files deleted from this directory while the search is progressing: First the heartbeat file, and then the sub directories.
The search has not completed and this hangs the search job forever.
Is this part of Hunk to delete these files as they age? If so, what is the delete age and or setting to control it?
We are also seeing any non hunk file being deleted from this directory. Is Hunk deleting them?
Overall – yes this is expected. Hunk deletes the files under the Hunk HDFS working directory.
What is not expected, is that it will delete these files before the job is completed and these files are still needed
How does the directory structure look under /mapr/tmp/dispatch? Each search job should have its own directory...
Do you have two instances of Hunk pointing to the same Hunk working directory in HDFS?