[MAHOUT-1669] Mahout nightly build lasts too long

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Won't Fix
• Fix Version/s: 0.11.2
• Assignee: Stevo Slavić
• Reporter: Stevo Slavić
• Created: 02/Apr/15 00:19
• Updated: 15/Mar/16 03:12
• Resolved: 15/Mar/16 03:12

Description

Mahout nightly Jenkins build job can last longer than 60min and then it gets automatically aborted by Jenkins.

Here is example from a recently failed build (#1881):

Build timed out (after 60 minutes). Marking the build as aborted.
[INFO] ------------------------------------------------------------------------
[INFO] Reactor Summary:
[INFO] 
[INFO] Mahout Build Tools ................................ SUCCESS [14.261s]
[INFO] Apache Mahout ..................................... SUCCESS [5.315s]
[INFO] Mahout Math ....................................... SUCCESS [2:36.064s]
[INFO] Mahout HDFS ....................................... SUCCESS [18.617s]
[INFO] Mahout Map-Reduce ................................. SUCCESS [21:15.011s]
[INFO] Mahout Integration ................................ SUCCESS [1:23.923s]
[INFO] Mahout Examples ................................... SUCCESS [5:15.152s]
[INFO] Mahout Release Package ............................ SUCCESS [9:43.293s]
[INFO] Mahout Math Scala bindings ........................ SUCCESS [2:58.073s]
[INFO] Mahout Spark bindings ............................. SUCCESS [5:09.176s]
[INFO] Mahout Spark bindings shell ....................... SUCCESS [21.001s]
[INFO] Mahout H2O backend ................................ FAILURE [7:29.662s]
[INFO] ------------------------------------------------------------------------
[INFO] BUILD FAILURE
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 57:07.897s
[INFO] Finished at: Thu Apr 02 00:00:25 UTC 2015
[INFO] Final Memory: 71M/772M
[INFO] ------------------------------------------------------------------------
channel stopped
Build was aborted
Finished: ABORTED

We should investigate and try to bring the build time down.

Some initial thoughts:

• MR module build time dominates (1/3 of build time); we should check what is done there, is it necessary and can something (e.g. tests) be parallelized
• There are few large artifacts produced and deployed/published by this build job; we should check if all of them are really necessary, and for necessary ones if we can reduce the artifact size to make publishing and build faster