rmi-cli v3.4.0 Release
@data-mlep has released rmi-cli v3.4.0 which includes major update in serving and several minor updates to improve the testability and maintainability of our ML platform.
Change logs
Major
- Migrate serving helm chart to use CDE managed charts to standardize our deployment in data team. The migration also enables these features:
- Enable workload identity for deployed services, which improve our security since we don’t need to create service account keys.
- Enable securing the service via Single Sign-On (SSO). One common use case for this is to secure swagger for the API docs. You can refer to this doc on how to enable this feature.
- Improve KFP installation to use installation managed by terraform. We also managed to create a more human recognizable URL instead of machine generated one. You can access the new KFP dashboard via https://kfp.mlplatform-prod.tvlk-data.com
- Remove airflow job support from
scheduler
Minor
- Migrate from KMS to central secret manager to follow the guideline set by Cloud Infra team and CDE
- Refactor
rmi-cli to improve testability
- Create unit test for commands module. The test coverage is ~90%
- Merge commands under
pipeline with scheduler
What RMI users need to do
- Since we will using CDE charts, we need to redeploy all services, which will introduce downtime to the service (we need to delete the old deployment, then deploy the service to use CDE chart). The schedule for this operation can be seen in this spreadsheet. If you have any concerns related to the schedule, please update the schedule in the spreadsheet before Fri, Oct 7th 2022. By default, we set the schedule as follows:
- Staging: Mon, Oct 10th, 7 am
- Production: Wed, Oct 12th, 7 am
- After the CDE charts migration has been done, please update your
rmi-cli in your local computer and CI/CD pipeline by executing rmi update -v v3.4.0. Please update the version before Oct 28th as we need to decommission the KMS as suggested by Cloud Infra 
:bows:
Reference
For more detailed informations, please refer to these documentation and guideline:
Support
Got any questions? Feel free to contact @data-mlep via #rm-support slack channel
Thanks! :bows: