From 3564859f037b2d125785956c38668950145897d9 Mon Sep 17 00:00:00 2001 From: david Date: Wed, 28 May 2014 04:24:31 -0700 Subject: [PATCH] Simplify Uploads by only sending Pass results Change-Id: Iffa95584bd3417391704707811cc9ef21527dc60 storyboard: https://storyboard.openstack.org/#!/story/108 --- ...y-uploads-by-only-sending-pass-results.rst | 129 ++++++++++++++++++ 1 file changed, 129 insertions(+) create mode 100755 specs/approved/simplify-uploads-by-only-sending-pass-results.rst diff --git a/specs/approved/simplify-uploads-by-only-sending-pass-results.rst b/specs/approved/simplify-uploads-by-only-sending-pass-results.rst new file mode 100755 index 00000000..0e93011e --- /dev/null +++ b/specs/approved/simplify-uploads-by-only-sending-pass-results.rst @@ -0,0 +1,129 @@ +============================================= +Simplify Uploads by only sending Pass results +============================================= + +Blueprint: https://blueprints.launchpad.net/refstack/+spec/pass-only-uploads +Storyboard: https://storyboard.openstack.org/#!/story/108 + +As part of helping the community accept the publication of results, refstack +test uploads should default to only PASS results. We are NOT uploading skips +or errors. This aligns with the interop objective because we want have a +positive assertion. + +Problem description +=================== + +Because fail results often include sensitive information it has been a pain +point for some of our early adopters to upload results to a foundation +controlled database. Since refstack really only cares about the things that +pass, we'll just parse the results and leave the fails out. + +Proposed change +=============== + +This would involve using the subunit parsing code inside the tester. We would +run the parser on the subunit before uploading the results to the api. + +We'll want to have a non-default option to send all data. Because I am sure +that some folks will want to use refstack internally for real debugging and +test regression. + +If a user tries to submit results to the public api server (Regardless of what +the non-defaulting option is set to.). It will either produce an error (which I +don't think is as useful) or it will just scrub the results anyways. + +Alternatives +------------ + +We could take the full upload and do the processing server side to accomplish +the DefCore objective; however, that does not avoid the data leak issue. + +Data model impact +----------------- + +None + +REST API impact +--------------- + +None + +Security impact +--------------- + +This change addresses security concerns. as far as I can tell it will not +create any new ones. + +Notifications impact +-------------------- + +It should mention that results are being scrubbed in the on screen log messages. +so that it's clear on every action. + +Other end user impact +--------------------- + +None + +Performance Impact +------------------ + +This will slow down the tester but not by a lot. However it will also speed up +the upload of the results since they will be trimmed. + +Other deployer impact +--------------------- + +None + +Developer impact +---------------- + +This could potentially be a maintance problem moving forward as we move the +subunit parsing utils into tempest then move the tester into its own repo. + +It will require that someone stays on top of things durring those changes to avoid +duplication of code. + + +Implementation +============== + +Assignee(s) +----------- + +Primary assignee: + dlenwell + +Other contributors: + catherine wrote the parsing utils. So some support might be needed. + +Work Items +---------- + +The tester already has a stubbed out function for this code. Just needs +to be filled in. + +Dependencies +============ + +None + +Testing +======= + +None + +Documentation Impact +==================== + +It should be written down that we aren't uploading fails. So its known to +operators and they don't have to be concerned about security leaks. + +But outside of that I don't see the need for documentation. + + +References +========== + +N/A \ No newline at end of file