What are you thinking to: Build time, stability....?
I have seen work to optimize the download of spark.
Travis starts from a clean env which is good but things take time to be
downloaded and setup.
I guess we have to log each failure which should not occur and tackle
them. I have seen multiple times such errors which should not occur:
Failed tests:
ZeppelinSparkClusterTest.pySparkDepLoaderTest:167->getSparkVersionNumber:220
expected:<FINISHED> but was:<ERROR>
ZeppelinSparkClusterTest.pySparkAutoConvertOptionTest:112->getSparkVersionNumber:220
expected:<FINISHED> but was:<ERROR>
ZeppelinSparkClusterTest.basicRDDTransformationAndActionTest:81
expected:<FINISHED> but was:<ERROR>
ZeppelinSparkClusterTest.pySparkTest:90->getSparkVersionNumber:220
expected:<FINISHED> but was:<ERROR>
ZeppelinSparkClusterTest.zRunTest:152 expected:<FINISHED> but was:<ERROR>
(https://s3.amazonaws.com/archive.travis-ci.org/jobs/111732702/log.txt)
Apparently checkIfServerIsRunning is the cause - maybe the server takes
time to open the socket...
On 08/03/16 19:49, Felix Cheung wrote:
I think there are rooms to optimize much more - feel free to propose or
contribute!
On Tue, Mar 8, 2016 at 10:29 AM -0800, "Eric Charles" <e...@apache.org> wrote:
Makes sense. No need to run tests if a module fails on compilation,
format...
On 08/03/16 18:52, Felix Cheung wrote:
Right now Travis does a full build first and then run the tests in a separate
steps.
On Tue, Mar 8, 2016 at 2:36 AM -0800, "Eric Charles" <e...@apache.org> wrote:
Hi there,
I see tests are disabled on Travis.
e.g. From
https://s3.amazonaws.com/archive.travis-ci.org/jobs/114476198/log.txt
$ export PROFILE="-Pspark-1.5 -Phadoop-2.3 -Ppyspark"
$ export BUILD_FLAG="package -DskipTests"
Is this on purpose?