Cumulative Update 1


This section introduces issues fixed in EnOS 2.3 CU1.

Core

Issue

Fix / Enhancement

Impact (if any)

Users can view the alert event but not the log context after setting the log alert.

Adjust the filter logic to support only Log filtering in the log metrics configuration.

None

IAM

Issue

Fix / Enhancement

Impact (if any)

Local storage security issues.

The key associated with the token in the local storage is placed in the cookie to prevent token theft.

None

IoT Hub

Device Integration Service

Issue

Fix / Enhancement

Impact (if any)

The input msg for Lookup Mapping must be an array, which can cause issues such as when mapping one to many, the array of values mapped does not follow the order in the CSV file.

The input msg can be in JSON or an array.

None

The HTTP Client and Refresh Token nodes will automatically add the Authorization field in the header when Anonymous is selected as the Authentication Method. If there is an Authorization field in the user’s request, it will be overwritten.

The Authorization field will not be added in the header.

None

The timeout is calculated in seconds instead of milliseconds for the HTTP Client and Refresh Token nodes.

The timeout is calculated in milliseconds.

None

When there are multiple entries for a node, the downstream nodes will have missing attributes.

Back up the attributes of each node before processing.

None

The metadata deserialization for HTTP Client and Refresh Token nodes will fail when caching at site and storing tokens, which will cause the node to report an error.

The serialization method is modified.

None

Enterprise Data Platform

Stream Processing Service

Issue

Fix / Enhancement

Impact (if any)

On the Pipeline Designer page, when click New Pipeline-Import, if the user imports an incorrect configuration (for example: the template of the configuration is not installed or does not exist), an error will be reported, and the original user pipeline will be accidentally deleted.

Enhanced the verification to ensure that the user pipeline will not be incorrectly overwritten when the parameter verification fails. And fixed the transaction issue with the interface of importing the configuration.

None

When the IAM backend service verifies whether the user login information has expired, if the authentication interface is abnormal, the error code 30000 will be returned, causing the user to log out.

Optimized the logic so that logging out only happened when the user’s login information had expired, and return a tip indicating that the query of user information failed in other cases.

None

When the user performs the Stop operation on a large number of (more than 100) pipeline tasks at the same time, or when the Stop operation is repeatedly retried due to a system abnormality, the streaming state management service will fail to submit the Kafka Offset successfully, resulting in the pipeline has been in the Stopping state.

Optimized the Kafka Consumer parameters of the stream processing state management service.

None

Time Series Data Management

Issue

Fix / Enhancement

Impact (if any)

On the Data Insight page, when the data points are negative and very close, the data points on the chart axes will be obscured.

Limits the maximum and minimum values of the y-axis of the chart to include the range of all data.

None

  • Get Asset Latest Data API query result contains localtime field by default.

  • Service error when Get Unformatted Policy API is called.

  • Get Asset Latest Data API query result no longer contains localtime field by default.

  • Fixed the calling logic of Get Unformatted Policy API.

None

  • Only a single kdc service is configured for kerberos authentication, so there are HA problems.

  • Tsdb-writer has thread safety issues in table building in some cases.

  • Changed how tsdb-queryengine and tsdb-writer-krondb kerberos are configured.

  • Changed the way to generate timestamps for table building of tsdb-writer-krondb to avoid thread safety issues.

None

Data Federation

Issue

Fix / Enhancement

Impact (if any)

When accessing MySQL through a single data source channel, if SQL is renamed by using as, an error will be reported when querying.

After the fix, you can use as to rename the query field normally.

None

Azure official report stated that Blob SDK versions below 12.14.1 have security vulnerabilities and need to be upgraded.

Security bugs have been fixed after the upgrade.

None

During data download, if the file size exceeds 4G, the download will fail. The failure message is “The zip compression algorithm cannot compress files larger than 4G”.

No limit on file size downloaded after fix.

None

During data download, if the parameter length of the file renaming field exceeds 2097152 bytes, the download will fail.

No limit on parameter length after fix.

None

Enterprise Analytics Platform

EAP Dashboard

Issue

Fix / Enhancement

Impact (if any)

The system does not respond when a user not in “eap_admin” group modifies the user group information in Permission Mgmt.

Only users in “eap_admin” group can modify user group information, otherwise a “no access” message prompts.

None

Dataset Management

Issue

Fix / Enhancement

Impact (if any)

When you enter an invalid dataset name, the error message does not provide clear information.

An informative error message appears when you enter an invalid dataset name.

None

If a dataset is created from Hive data source with HDFS Connection enabled, you can only view up to 1000 data in the Jupyter Notebook by invoking the dataset code.

You can now view all data by invoking the code to Jupyter Notebook.

None

MI Lab

Issue

Fix / Enhancement

Impact (if any)

Failed to run a pipeline if NotebookEx/PythonEx/ShellEx operators include references to files in the internal storage with spaces in the filename.

NotebookEx/PythonEx/ShellEx operators now supports referencing to files with spaces in filenames.

None

MI Hub

Issue

Fix / Enhancement

Impact (if any)

If a model is deployed via operators, the deployment status remains “deploying” when switching model versions.

You can now successfully switch model versions.

None

When a model deployment instance is switched from Canary Deployment to Blue/Green Deployment, you cannot request feedback interface, where an error “jsondata is not jsonstruct” appears.

You can now request the feedback interface after switching deployment methods.

None

The model data displayed the line chart of Model Indicator Monitoring page is incorrect after a service calling test or a feedback test.

The line chart now displays the correct model data in Model Indicator Monitoring page.

None

For models deployed via Blue/Green Deployment, an error “data is not json struct” occurs when requesting feedback interface after calling model service.

You can now request feedback interface successfully on the model deployed via Blue/Green Deployment.

None

If the model is staged via a Docker/Mlflow/S2i/Thirdparty Model Version Register operator, the size of model files are displayed as 0 kb in Model Deployment page when importing model files from Git Source.

Correct file size is displayed now.

None

For a model version staged via an operator, the branch/tag information of the model files is missing in Model Deployment page if the model files are imported from Git Source.

You can now view the branch/tag information of models in Model Deployment.

None

If a model is staged via an operator, the model information is missing when you copy an existing version.

The model information is carried over to the new version when you copy an existing version.

None

If a model version is staged via Docker Model Version Register operator, an error “model build failed” appears when you copy the model version.

You can now successfully copy the model version staged via Docker Model Version Register operator.

None

Blue/Green Deployment does not support Object in List structure.

Blue/Green Deployment supports Object.

None

Blue/Green Deployment only returns the prediction outputs of the latest model version.

Blue/Green Deployment supports recording the outputs of all model versions.

None

MI Pipelines

Issue

Fix / Enhancement

Impact (if any)

If there are two Pipeline Trigger operators in a DAG graph, you cannot view the operator details by clicking one of the operators.

You can now view the details by clicking an operator in a DAG graph.

None

When you add a scheduling configuration for a pipeline from the experiment list, the pipeline version scheduled to run is not the version you specified.

The schedule now runs on the specified pipeline version in Scheduling Configuration.

None

No logs are recorded when model prediction operators fails to predict in models deployed via Blue/Green Deployment.

Error reports are recorded in the logs when prediction operators fails.

None

Operators in a sub-canvas cannot reference to the output of an operator in the main canvas.

MI Pipeline supports nested multi-level references.

None

After specifying values for workspace/entrypoint/requirement parameters in NotebookEx/PythonEx/ShellEx operators, you cannot view the full directory path of these parameters.

You can now view the full path of the parameters in NotebookEx/PythonEx/ShellEx operators.

None

In the wind-power-forecast sample pipeline, “error - modelversion test failed” and “error - output data server may not work now” occur if the environment resource is insufficient when staging the model version.

The value of “enforce_register” parameter is now “true” by default in Mlflow Model Version Register operator to ensure that the model can be staged with limited resources.

None

The wind-power-forecast sample pipeline does not support exporting pipeline results to Hive data source.

  • The prediction results can now be saved to EnOS Hive, which enables the visualization in DTV.

  • Pipeline orchestration is simplified.

None