[
http://issues.apache.org/jira/browse/HADOOP-477?page=comments#action_12435036 ]
arkady borkovsky commented on HADOOP-477:
-----------------------------------------
As Owen pointed out few weeks ago, all what is necessary is to provide a
standard environment for the executables.
I suggest 3 things to be done:
(1) privide a "standard", meaningful, and documented environment to commands
executed by streaming.
It should include at least
-- PATH that has /usr/bin, etc. -- so that commands like grep, perl,
python, awk etc are found, as well as hadoop uitilities
-- LD_LIBRARY_PATH (to get to C and C++ shared libraries)
(2) allow the user to provide a file (similar to .cshrc) to define environment
variables.
(3) allow the user to define environment variable in the Streaming command.
The (1) is probably a must.
I strongly prefer (2) over (3), as I would just borrow someone set up that
works, and I do not know what are the actual necessary setting.
> Streaming should execute Unix commands and scripts in well known languages
> without user specifying the path
> -----------------------------------------------------------------------------------------------------------
>
> Key: HADOOP-477
> URL: http://issues.apache.org/jira/browse/HADOOP-477
> Project: Hadoop
> Issue Type: Bug
> Components: contrib/streaming
> Reporter: arkady borkovsky
>
> If the executables for -mapper or -reducer are well-known (grep, cat, awk),
> Streaming should make sure that the executable is found.
> If a script for -mapper or -reducer are in a well-known language (.pl, .py),
> Streaming should execute it with the correct language processor.
> Reason:
> many jobs get started from machines with a different environment from that on
> the cluster.
> On another hand, different clusters may have different environments.
> Also, a user may have no access to the cluster machines.
> Because of this, a user may be unable to specify correct paths for standard
> commands, and correct language processors for scripts.
> Implementation:
> Stream may tailr the commands by prepending the path, or the name of language
> processor.
> Another solution is to make sure that the commands are executed in a
> "meaningful" environment (with good $PATH, and other variables Unix users are
> accustomed to count upon).
> Once again, Streaming is user facing tool -- it is not a library or a
> hackable example that the users are to modify for their needs. So it should
> work out of the box.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
