Multiple jobs #55
Multiple jobs #55
Conversation
sesheta
added
the
size/L
|
Build failed.
|
|
Build failed.
|
|
Build succeeded.
|
| create_inspect_buildconfig(_OPENSHIFT, inspection_id, dockerfile, specification) | ||
| inspection_ids = [] | ||
| for _ in range(job_count): | ||
| inspection_ids.append(_generate_inspection_id()) |
There was a problem hiding this comment.
I'm not sure if we want it like this. Ideally we would like to have one ID per request to make only one build (we don't need to build the same thing multiple times). For this build then there should be executed N jobs. These jobs should be referenced with the same ID due to tractability so we can easily point to specific inspections from Thoth.
This can be accomplish by using labels - we mark a job with a label that is set to inspection_id so we can query OpenShift API for these ids. The actual job names can then be numbered like <inspection_id>-0, <inspection_id>-1, ...
There was a problem hiding this comment.
in general, this is a #techdebt and ok for a POC, but we should not forget to redesign this using the Operator pattern
There was a problem hiding this comment.
Maybe I'm missing something here - how will operator be helpful here?
There was a problem hiding this comment.
We have a restful interface (CRD), and logic that is managing things like BuildConfig and ImageStream (secondary resources), this logic adjusts parameters based on an observed cluster status (#38), and Builds etc need to be cleaned up after they are not longer in use. Sounds like the perfect description of a Kubernets-native Application, thus pattern to implement this is an Operator. Not saying we need to do it this year, but we should keep it in mind.
There was a problem hiding this comment.
I'm not sure if we want it like this. Ideally we would like to have one ID per request to make only one build (we don't need to build the same thing multiple times). For this build then there should be executed N jobs. These jobs should be referenced with the same ID due to tractability so we can easily point to specific inspections from Thoth.
This can be accomplish by using labels - we mark a job with a label that is set to
inspection_idso we can query OpenShift API for these ids. The actual job names can then be numbered like<inspection_id>-0,<inspection_id>-1, ...
Havent solved this as of now. Can we merge this as of now as POC ? and then implement as needed
|
Build succeeded.
|
amun/core.py
Outdated
| @@ -137,23 +134,20 @@ def create_inspect_job(openshift: OpenShift, image_stream_name: str, specificati | |||
| template = response['items'][0] | |||
|
|
|||
| if 'run' in specification: | |||
| if 'requests' in specification['run']['requests']: | |||
|
@ace2107 thanks for the implementation. I would postpone this one a little. For now we are OK with one job submission to see end2end flow. We will definitely need this later, but ideally referenced with one inspect id and one build per inspection type. WDYT? |
|
Yes, we can have one build and job for now to make sure it is working end to end and later have it with -0,1,2... That will just be a slight change in the code , and can be implemented now too if needed |
|
Build failed.
|
|
Build failed.
|
Solving issue #38
If this looks good the InspectionResponse description can be deleted too