Re-Publishing Decisions using the same name, in SAS Intelligent Decisioning, does not end

[JAE2019]

Hi Experts! We're currently working with SAS Intelligent Decisioning running in LTS 2022.1, and we are having challenges with publishing Decisions in MAS. Sometimes, when we re-publish a decision flow - publishing does not end. At the onset, increasing the RAM helped resolve the issue. But then we experience the same problem again (While we maintained the same increased RAM). I say "Sometimes" because re-publishing will work for sometime for a specific decision_flow name, until such time we cannot publish it again and we implement the work-around of publishing it under a new name.

We have been looking at MAS logs, also sending to Technical Support. These are the last two lines of the log as reference recently. So far we've only seen

• the status of publishing to be stuck in PENDING: moduleId=DECISIONFLOWMAIN5_5, state=PENDING, errors=[], toString()=AbstractModifiableResource[createdBy=dev_user, creationTimeStamp=2023-02-15 01:52:52.214, modifiedBy=dev_user, modifiedTimeStamp=2023-02-15 01:52:52.214
• and another one on ACCESS_TOKEN_FAILURE which we're unsure if related since the team has not been experiencing any access token related error during this time : [ACCESS_TOKEN_FAILURE] Could not obtain access token.\nerror=\"access_denied\", error_description=\"Access token denied

We are trying to understand what is the key root cause. Attached are the last few lines of the log for reference.

1) Has anybody experienced the same, and how have you addressed this problem?

2) Is there any "un-publish" function in any SAS CLI/ APIs, or anywhere in the GUI (that we are supposed to use before re-publishing the same flow? So far we have only seen the APIs for listing the published flows and deleting published flows, and none from the SAS Viya Application.

3) Is publishing really expected to be that RAM intensive? The more complex the decision flow, the higher RAM resources is taken during publishing - even before execution?

4) What process actually happens as we publish the decision flow to MAS? Does it somehow execute the script/logic such that it tries to issue an actual connection to the databases and APIs that are being called by the code files within the Decision Flow?

5) What other elements/components, that may influence the publishing process, can you recommend for us to look at?

Thank you!!!

[alexal]

I would increase App.tk.MAS logger to the debug level and see what else will be in the log. See the following page
https://go.documentation.sas.com/doc/en/mascdc/v_021/masag/n1er2r63lzub91n1b2o174bbgp6g.htm. In addition to that, have you checked the model publishing pod for any errors?

[JAE2019]

Thank you @alexal. We have activated DEBUG, but the two lines I have shared are the only errors popping out.

How to we go about checking the model publishing pods?

[alexal]

kubectl logs -f sas-model-publish-*****-****

[JAE2019]

Thank you!! we will try this out @alexal !!