incubator-blur-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chris Rohr <>
Subject Re: Testing weirdness
Date Fri, 26 Sep 2014 11:18:59 GMT
After rebasing from master here are the steps I followed:

1. (from root) mvn -Dhadoop2-mr1 -U -DskipTests clean install
2. (from blur-console) mvn -Dhadoop2-mr1 test


ERROR MiniCluster - File
file:apache-blur-0.3.0-incubating-SNAPSHOT/commands does not exist

FATAL Config - Unable to start in dev mode because MiniCluster isn't
in classpath

I have pushed my branch up to git.


On Thu, Sep 25, 2014 at 10:36 AM, Tim Williams <> wrote:
> On Thu, Sep 25, 2014 at 7:47 AM, Chris Rohr <> wrote:
>> This isn't totally a Blur question exactly, but I'm having some
>> weirdness running console tests and I thought I would get some insight
>> to see if anyone has any ideas.  Here are the scenarios that I have
>> performed an the outcomes (note, this is on a feature branch I'm
>> currently working, I will indicate the scenario outcomes on current
>> master as well).
>> #1 Running each test class standalone in Eclipse
>> #1 Outcome (Pass): each test passes
>> #2 Running all tests in Eclipse
>> #2 Outcome (Errors): The first test usually passes and then all
>> subsequent tests fail with connection loss to Zookeeper (from
>> minicluster).  Note in this scenario, each test class starts the mini
>> cluster and then after shuts it down
>> #3 Running the ConsoleTestSuite class in Eclipse
>> #3 Outcome (Failures): MiniCluster starts before all, the tests run
>> but any assertion after changing Blur (i.e. creating a table or
>> querying) fails like the action never happened, then the MiniCluster
>> shuts down.
>> #4 Running each test class standalone in Maven
>> #4 Outcome (Errors): MiniCluster isn't in the classpath (fyi, the
>> blur-core module is a test dependency not compile scope) ----- This
>> passes in current master
>> #5 Running all tests from blur-console directory in maven
>> #5 Outcome (Errors): Same as #4 ----- This passes in current master
>> #6 Running the ConsoleTestSuite in Maven
>> #6 Outcome (Errors): Same as #4 ----- This has the same affect in
>> current master as #3
>> #7 Running all tests from root in maven
>> #7 Outcome (Errors): Never get to blur-console as other modules fail first
>> I'm sure this is something I'm just missing.  To give some more info
>> on the changes I made on my feature branch.  I am trying to remove the
>> dependency on ZookeeperClient hence removing the dependency on
>> blur-core for the console.  I did have to leave it in the test scope
>> (as well as the blur-core test-jar) for the MiniCluster.
> Have you rebased the branch with master recently?  I remember there
> was some issue with us using Zookeeper in some places instead of our
> ZookeeperClient - ours has retry logic built in so it gets around some
> timing issues.  Is #6 the only one that you can reproduce in master -
> can you give a stack trace or more details on it?  If you wanna push
> your branch, if it's something your doing, we should be able to help
> spot it in the diffs pretty quickly?
> --tim

View raw message