solved in thread "select count(1)" on Mon, 03 Jan
Hello Carl and Dilip,
yep, was the HADOOP_HOME. The jars and classpath are allright. I
didn't realize that eclipse ignores user specific system vars (under
osx). I solved it with this:
http://efreedom.com/Question/1-829749/Launch-Mac-Eclipse-Environment-Variables-Set
Thank you
Malte
2011/1/3 Car
Hi Malte,
This is the root of your problem:
> java.io.IOException: Cannot run program "null/bin/hadoop": error=2, No
such file or directory
If you look at MapRedTask.execute() you'll see that the location of the
hadoop script is determined based on the value of the hadoop.bin.path
configuration
Another possibility is that some required jars are not in your classpath
when running your JDBC client. That could explain why everything works from
the Hive CLI and HiveServer.
I had written an article on using Hive from Jython -
http://csgrad.blogspot.com/2010/04/to-use-language-other-than-java
To specify things: I run the embedded mode with the url:
"jdbc:hive://". If I take "jdbc:hive://localhost:1/default" with
the hiveserver mode, the error will not occur and everything's fine.
But I need multithreated access to my data.
Malte
2011/1/3 kiteraddict :
> Hi Dilip,
> Unfortunately, n
Hi Dilip,
Unfortunately, not. No special characters. Tried a symlink as well...
The Path seems ok, and the strange is: select count(1) from the cli
woks.
I get the same error in Eclipse and in Netbeans (OSX 10.6.4). However
thank you for your tip!
Malte
2011/1/3 Dilip Joseph :
> Does the full path
Does the full path to your hadoop installation contain a '+' character? I
ran into a similar problem where a bug in the DataNucleus libraries used by
hive prevented it from constructing the full path to your hadoop
installation, if the path contained a '+'. The solution was to remove the
'+' from
see more specific question at Mon, 03 Jan, 13:55
