Exactly. IMO, that would cut the user experience pain hugely. With that
fixed users can run smoke tests out of the box just like running
docker-provisioner.

2016-01-16 10:14 GMT+08:00 Konstantin Boudnik <[email protected]>:

> On Fri, Jan 15, 2016 at 06:13PM, Evans Ye wrote:
> > Yeah. It's quick&dirty to add SPARK_HOME into smoke-test.sh, but I think
> > it's better to have those variables default exported in gradle before
> > launching smoke-tests. That way people won't need to go to
> >
> https://cwiki.apache.org/confluence/display/BIGTOP/Running+integration+and+system+tests
> > everytime running smoke-tests.
> > Do you think it's a good way to go?
>
> You mean like what I've done for BIGTOP_HOME? Yeah, I think we certainly
> can
> push these defaults into the gradle system from the script. Which makes
> sense
> as we are running smokes from Gradle anyway. If this is what you had in
> mind -
> I am all for it. I have opened up  BIGTOP-2250 and will try to work on it
> next
> week unless someone takes it on ;)
>
> Cos
>
> > 2016-01-15 12:24 GMT+08:00 Konstantin Boudnik <[email protected]>:
> >
> > > Hey guys.
> > >
> > > Looks like moving the smokes into the top-level had an interesting
> effect
> > > on
> > > the provisioner. BIGTOP-2244 explains the issue which was failing our
> CI
> > > smoke
> > > testing. Now all is back to blue again. However, looking into the
> > > provisioner
> > > jobs, like
> > >
> > >
> > >
> http://ci.bigtop.apache.org/job/Bigtop-Docker-provisioner-centos-6/80/console
> > >
> > > I've noticed that soma of the tests arent running, and looks like there
> > > never
> > > were ran, because:
> > >
> > >
> > > Build file '/bigtop-home/bigtop-tests/smoke-tests/build.gradle' line:
> 26
> > >
> > >
> > > * What went wrong:
> > > Execution failed for task ':bigtop-tests:smoke-tests:spark:test'.
> > > > undeclared env variable: SPARK_HOME
> > >
> > > And clearly, our bigtop-deploy/vm/utils/smoke-tests.sh isn't setting up
> > > SPARK_HOME variable. Shall it be fixed or there some other plans for
> it?
> > >
> > > Cos
> > >
> > >
>

Reply via email to