Tibco

TIBCO ModelOps 1.3.0 Release Notes

These are the release notes for TIBCO® ModelOps 1.3.0.

Contents

Release Notes

The release notes list the major updates in the 1.3.0 release of TIBCO® ModelOps. In addition, it also includes the migration procedures and compatibility matrix.

New Features

The following new features have been added in this release of TIBCO® ModelOps.

Key Summary
TMO-3906 TIBCO ModelOps now supports configurable environment names.
TMO-4226 User Interface now provides the support for R dependencies.
TMO-4266 TIBCO ModelOps now supports OIDC authentication in the Jupyter integration.
TMO-4282 TIBCO ModelOps uses Python 3.10 instead of 3.8, as in the previous version.
TMO-4304 TIBCO ModelOps now supports deployment to Streaming clusters outside of K8s.
TMO-4309 Users can now access external scoring services from TIBCO ModelOps via a scoring service.

Migration and Compatibility

Migration and compatibility are not affected in this release of TIBCO® ModelOps.

Resolved Issues

The following issues are resolved in this release of TIBCO® ModelOps:

Key Summary
TMO-3743 While deploying a flow through pipeline, date and date-time type fields are now supported .
TMO-4321 Users can now create and update their Scoring Flows with the new field in the Model Input Schema without reloading.

Known Issues

The following issues exist in this release of TIBCO® ModelOps:

Key Summary and Workaround
TMO-2562
  • Summary: TIBCO ModelOps does not support Avro unions and the default Avro schema writers in Java, Pyspark, and Scala write out string types as unions of string and null.
  • Workaround: None.
TMO-2676
  • Summary: Filter for the usage of scoring pipeline does not give correct data.
  • Workaround: None.
TMO-2735
  • Summary: The current Kafka channel source and sink implementations are based on EventFlow applications. The EventFlow type system does not have a pure date time. Hence, the date data type is converted to datetime format in sink.
  • Workaround: None.
TMO-3337
  • Summary: The Python Runner does not support the Python syntax from <module-name> import <name> for modules in the requirements.txt file. This does work for standard modules, for example: from typing import Iterable.
  • Workaround: Use import <module-name> and then specify the complete package name when referenced.
TMO-3420
  • Summary: Changing a user's roles requires the TIBCO ModelOps server to be restarted to take affect.
  • Workaround: None.
TMO-3460
  • Summary: Scoring pipelines do not automatically reconnect to their associated data channels on network errors. This prevents any more data from being processed by the scoring pipeline following the network error.
  • Workaround: Shutdown and restart the scoring pipeline.
TMO-3733
  • Summary: While approving a data channel or pipeline artifact to the Testing or Production environments, there are two requirements: the artifact must be published and the artifact must be approved to any preceding environments in the predefined environment flow. Currently informational messages only specify the requirement to be published.
  • Workaround: None.
TMO-3735
  • Summary: When exporting projects with model artifacts, the Model Type property is not being included in the exported data.
  • Workaround: When importing the project, a user must manually set this property again on their model, or set the default value to Other.
TMO-3911
  • Summary: Following a fresh installation, TIBCO ModelOps Server may not connect to the Metrics Server. This will cause the Performance tab to display no data.
  • Workaround: Restart the Metric Server Pod
TMO-4150
  • Summary: LIQO cannot be installed on EKS since it does not support STS currently.
  • Workaround: None.
TMO-4572
  • Summary: Scoring flows are currently not validated until deploy-time. An invalid scoring flow can be deployed (and published) with errors. This scoring flow will fail when deployed.
  • Workaround: Ensure that a scoring flow does not have any identified errors before deploying or publishing.
TMO-4579
  • Summary: When restoring packages from a lock file in TIBCO ModelOps, some R packages automatically brings in transitive package dependencies and recursively install a large number of R packages that are not specified in the lock file. This slows down the installation of model package dependencies when loading R models in TIBCO ModelOps.
  • Workaround: Use an alternate R package that offers a similar model creation capabilities but might install a relatively less number of transitive package dependencies.