> On Aug. 25, 2015, 3:31 a.m., Maxim Khutornenko wrote:
> > src/test/sh/org/apache/aurora/e2e/http/http_example.aurora, lines 45-48
> > <https://reviews.apache.org/r/37742/diff/1/?file=1049818#file1049818line45>
> >
> >     I am a bit concerned we will no longer have any constraint coverage in 
> > our e2e tests. Can you clarify why post deploy RPM env altering is not an 
> > option here?
> 
> Bill Farner wrote:
>     It's certainly an option, but testing against vanilla setups lowers the 
> barriers pretty considerably since things are likely to be 
> organized/configured differently on different platforms.  I'm not firm on 
> this change, but i do find the point of constraint coverage a weak one 
> without an explicit test case for it (rather than the current repetition 
> which is really a relic of old behavior).

I see your point about lowering testing barriers. At the same time, it feels 
like having an artificial restraint on modifying slave config prevents us from 
adding constraint tests in either approach (e2e or RPM). I'd rather have an 
option to address lack of coverage later if needed.

Splitting RPM and e2e test configs seems like a better way forward to me. This 
will let us keep RPM tests simple and lean (i.e. just run a subset of e2e if 
needed) while not precluding custom mods needed to test specific features. One 
example: oversubscription tests will require adding somthing like this into the 
slave config, which may or may not be of value to RPM testing:
```
--resource_estimator=org_apache_mesos_FixedResourceEstimator 
--modules={"libraries":[{"file":"/usr/local/lib64/libfixed_resource_estimator.so","modules":[{"name":"org_apache_mesos_FixedResourceEstimator","parameters":[{"key":"resources","value":"cpus:8"}]}]}]}
```


- Maxim


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/37742/#review96286
-----------------------------------------------------------


On Aug. 25, 2015, 1:11 a.m., Bill Farner wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/37742/
> -----------------------------------------------------------
> 
> (Updated Aug. 25, 2015, 1:11 a.m.)
> 
> 
> Review request for Aurora and Maxim Khutornenko.
> 
> 
> Repository: aurora
> 
> 
> Description
> -------
> 
> These are no longer needed now that the scheduler isn't injecting these 
> settings by default for jobs.  It's beneficial for them to be removed, as it 
> will make it easier to point the e2e test at an arbitrary environment to 
> exercise it.  I would like to use that approach to test RPMs/debs.
> 
> 
> Diffs
> -----
> 
>   examples/vagrant/upstart/mesos-slave.conf 
> 2b6a60673fc0a7ea3b73471701cd5d3efd6ce639 
>   src/test/sh/org/apache/aurora/e2e/http/http_example.aurora 
> c1a10d8ea60be6aa56e4517fb34288d7d5ae1480 
>   src/test/sh/org/apache/aurora/e2e/http/http_example_docker.aurora 
> 870b3e68035fdf86253cf9b92b606645134b3369 
>   src/test/sh/org/apache/aurora/e2e/http/http_example_docker_updated.aurora 
> e55aad3a58d4e3c19332e06b70771f51f07aa9b7 
>   src/test/sh/org/apache/aurora/e2e/http/http_example_updated.aurora 
> 423dd4d4e8b03c2f852e25acd9340bd6288b7d24 
> 
> Diff: https://reviews.apache.org/r/37742/diff/
> 
> 
> Testing
> -------
> 
> e2e tests pass
> 
> 
> Thanks,
> 
> Bill Farner
> 
>

Reply via email to