JobTracker does not start when specifying a Windows path for hadoop.tmp.dir

[Holger Stenzhorn <ho...@ifomis.uni-saarland.de>]

Hi,

I am trying to get Hadoop running in the "pseudo-distributed 
configuration" on Cygwin using the latest SVN source but fail...
In the "hadoop-site.xml" I specify the variable "hadoop.tmp.dir" to have 
the value "c:\development\hadoop\tmp" (also tried with value 
"c:/development/hadoop/tmp").
(Besides that there are no changes to the default configuration.)
Looking at the various log files, the "namenode", "secondarynamenode", 
"datanode" all start fine (i.e. there are no exceptions).
But looking at the "jobtracker" log, I can see that this did not start 
properly (and hence the "tasktracker" log shows that it cannot connect):

2007-11-07 17:24:31,287 FATAL org.apache.hadoop.mapred.JobTracker: 
java.lang.IllegalArgumentException: Pathname 
/c:/development/hadoop/tmp/mapred/system from 
c:/development/hadoop/tmp/mapred/system is not a valid DFS filename.
    at 
org.apache.hadoop.dfs.DistributedFileSystem.getPathName(DistributedFileSystem.java:102)
    at 
org.apache.hadoop.dfs.DistributedFileSystem.delete(DistributedFileSystem.java:143)
    at org.apache.hadoop.mapred.JobTracker.<init>(JobTracker.java:686)
    at org.apache.hadoop.mapred.JobTracker.startTracker(JobTracker.java:120)
    at org.apache.hadoop.mapred.JobTracker.main(JobTracker.java:2054)

Cheers,
Holger