Sean Dague wrote:
On 05/07/2015 02:29 PM, Joshua Harlow wrote:
Boris Pavlovic wrote:
Sean,
Nobody is able to track and know *everything*.
Friendly reminder that Heat is going to be removed and not installed by
default would help to avoid such situations.
Doesn't keystone have a service listing? Use that in rally (and
elsewhere?), if keystone had a service and each service had a API
discovery ability, there u go, profit! ;)
Service listing for test jobs is actually quite dangerous, because then
something can change something about which services are registered, and
you automatically start skipping 30% of your tests because you react
correctly to this change. However, that means the job stopped doing what
you think it should do.
*This has happened multiple times in the past*. And typically days,
weeks, or months go by before someone notices in investigating an
unrelated failure. And then it's days, weeks, or months to dig out of
the regressions introduced.
So... test jobs should be extremely explicit about what they setup and
what they expect.
Makes sense, perhaps all the test (and/or test-like) frameworks could
share some code + common config that does this, seems to be something
simple (and something that all could use for pre-testing validation of
all the expected services being alive/active/up/responding...)?
^^ Just an idear,
-Josh
-Sean
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
