IntroductionBy default, versions earlier than mahout0.9 do not support hadoop2.2.0 and later versions. However, in many cases, because hadoop in the cluster environment is a version later than 2.2.0 and mahout must be used, you need to compile the source code, so that mahout supports hadoop2. Welcome to reprint, please indicate the source: http://blog.csdn.net/u010967382/article/details/39088035
Build wheelsMahout runs on hadoop2.2.0 with a wide range of requirements. Therefore, we must have already planted a tree. Note that not four update patches are used in order, but one is required! After reading the patch discussion in detail, I think the two patches are the most reliable: 1329-3.patch
1329-3-additional.patch
As tested by the author, mahout0.9, 1329-3-additional.patch cannot be successfully applied.
1329-3.patchIt can be applied successfully.
Step 1: patch the mahout source codeDownload the source code to the http://archive.apache.org/dist/mahout/ or https://github.com/apache/mahout/releases and unzip it to the server. Download 1329-3.patch to the https://issues.apache.org/jira/browse/MAHOUT-1354 and copy it to the server.
Run the following command in the source code root directory to send a patch: [email protected]: ~ /Mahout-distribution-0.9 $
Patch-P0 <../mahout0.9-patch-for-hadoop2/1329-3.patchThe patching file Core/POM. xmlpatching file integration/POM. xmlpatching file Pom. XML does not report an error and the upgrade is successful.
Step 2: Compile the source code for hadoop2.2.0Run the following command in the source code root directory to compile the source code for hadoop2.2.0 [email protected]: ~ /Mahout-distribution-0.9 $ MVN clean package-dhadoop2.version = 2.2.0
...... Long wait ......
Note:: I didn't add the-dskiptests parameter to understand the test results after the patch is created. The results show that all tests have passed, so there is no problem!
Verify
On the patch page, the author of the patch has the following dialog:
Sergey svinarchuk added a comment-21/FEB/14
I tested unit tests and examples with hadoop1 and hadoop2. all tests and examples passed.
I believe that Daniel's character and integrity, so do not test all examples, only test the Bayesian text classifier, see http://blog.csdn.net/u010967382/article/details/39088285
The entire process is normal!
[Ganzhou] mahout0.9 patch to support hadoop2.2.0