[minor] Integrate aiservice into mas-install pipeline

[Bhautik-Vala]

Enable mas-install pipeline to install AI Service as well - eventually combining aiservice-install pipeline into mas-install pipeline -> removing aiservice-install pipeline.

Story Link: https://jsw.ibm.com/browse/MASR-6486
Test story: https://jsw.ibm.com/browse/MASR-6487
Related PR:
ibm-mas/python-devops#141

Testing:

[durera]

At the moment we've basically attached the aiservice pipeline to the side of the existing MAS one, creating duplication in places ... what I'm looking for here is full integration into the pipeline

• sls - we are already installing sls in the pipeline, we shouldn't need a seperate stage, if customer is only installing aiservice, installing mas+aiservice, or just installing mas on it's own then the existing sls stage in the pipeline should be only place we "do sls stuff"
• dro - as above, dro is a cluster scoped install, we don't have a mas dro and an aiservice dro
• mongodb - same, our strategy is one mongodbcluster per ocp cluster, and everything that uses it uses it

I'd expect something like this to be what we are inserting:

• minio (e.g. run after ibm-catalogs)
• db2-aiservice (e.g. run after cert-manager)
• odh (e.g. run after ibm-catalogs)
• aiservice (run after existing sls, existing dro, db2-aiservice, odh)

Also, let's squash aiservice-tenant into a second step in aiservice task, see how I have done this for all mas apps, as we are trying to simplify the pipeline/make it easier for customers to read/understand/process.

[Bhautik-Vala]

@durera
Regarding separation of two similar tasks (e.g., sls/dro/mongo): when I tried to integrate the AIServices pipeline into MAS, the main issue was that when standalone AIServices is installed, we have to pass aiservice_instance_id as mas_instance_id to the SLS role. I thought a lot about how we could use a single task for both scenarios (i.e., standalone AIServices and with MAS). We want the dynamic nature of mas_instance_id’s value, but this can’t be achieved through Jinja templating because at runtime only we know whether mas_instance_id should be mas_instance_id or aiservice_instance_id for the SLS/DRO/Mongo role.

So I kept two tasks, but only one gets executed depending on the scenario.

When we install MAS + AIServices, this is the list of tasks:

Can you please suggest any tips on how we can dynamically set the value of mas_instance_id within a single task depending on the scenario?

Also, I will squash aiservice-tenant into the aiservice task.

[Bhautik-Vala]

@durera I have Integrate aiservice-tenant into the aiservice task.

[durera]

Can you please suggest any tips on how we can dynamically set the value of mas_instance_id within a single task depending on the scenario?

Assuming we are using mas_instance_id here to generate a "fake" slsfcfg, mongocfg, etc because that's just what the code in the aiservice provision expects then I suggest something like this (e.g. in the sls task):

  params:
    - name: mas_instance_id
      type: string
      description: Instance ID
      default: "" # By default, no config will be generated
    - name: aiservice_instance_id
      type: string
      description: AIService Instance ID - when installing AI Service in standalone mode we still need to generate the MAS slscfg object, because it is used to populate the AIService configuration
      default: ""

  stepTemplate:
    env:
      - name: MAS_INSTANCE_ID
        value: $(params.mas_instance_id)
      - name: AISERVICE_INSTANCE_ID
        value: $(params.aiservice_instance_id)

  steps:
    - name: sls
      script: |
        #!/bin/bash
        
        # Support AI Service standalone
        if [[ "$MAS_INSTANCE_ID" == "" ]] && [[ "$AISERVICE_INSTANCE_ID" != "" ]]; then
            echo "Setting MAS_INSTANCE_ID based on AISERVICE_INSTANCE_ID for standalone AI Service install"
            MAS_INSTANCE_ID=AISERVICE_INSTANCE_ID
        fi
        /opt/app-root/src/run-role.sh sls

This will make the task use aiservice-instance-id as a "fake" mas-instance-id when the role is ran, so that the MAS config resource file is still generated and can be used later in the pipeline.

[Bhautik-Vala]

Hi @durera
have done the changes regarding of sls/dro/mongo and tested for mas install and mas aiservice-install both here are the results:


can you please review again. Thank you.

[durera]

This is ready to merge ... approval pending the readiness of related PRs in python-devops and ansible-devops

[Bhautik-Vala]

This is ready to merge ... approval pending the readiness of related PRs in python-devops and ansible-devops

@durera For phase-1 we need only python-devops changes in this cli PR, right now python-devops is ready with required changes and I've tested that - can you please merge python-devops and this cli PR.