Deleting model versions

About this task

You cannot delete a deployed version of the model directly. If you want to delete a deployed version, you must first change the deployed version of the model, and then delete the original model.

Procedure

1. From the AI Model Management Training tab, click Log anomaly detection - natural language.

2. From the Log anomaly detection - natural language page, click the Model versions tab.

3. To delete a model version, go to the row of the version that you want to delete and click the options menu to expand it. Then, click Delete.

   Note: You cannot delete a deployed version of the model directly. If you want to delete a deployed version, you must first change the deployed version of the model and then delete the original model.

Resetting the log anomaly training environment

If you need to reset your log anomaly training environment, complete the following steps. Proceed with caution and do not attempt these steps on a production or critical Cloud Pak for AIOps environment.

1. Run the oc project command to set the context to the project where Cloud Pak for AIOps is deployed.

2. Important: Do not delete any of the following index files. Deleting these files might delete data that corresponds to other AI activities, such as metric anomaly detection training. However, you can use _delete_by_query to delete documents that are referenced by these index files.

   • trainingdefinition
   • trainingsrunning
   • trainingrun
   • trainingstatus

3. Specify the variables and delete the following index files. Deleting an index uses a cascading delete and deletes the documents that are referenced by the index.

   • Any 1000-1000-*-logtrain index files

   • The 1000-1000-vN-* or 1000-1000-v*-models index files of the particular version that you want to reset. Specify values for the asterisk (*) and N variables. The N variable is an integer number that indicates the version, such as in 1000-1000-v1-applications. If you want to delete all log anomaly model versions, use 1000-1000-vN-* or 1000-1000-v*-models.

   • The 1000-1000-log_models_latest index file

   • See the following command as an example:

     export ELASTIC_SET="$(oc get statefulset -l app.kubernetes.io/managed-by=ibm-elasticsearch -o jsonpath='{.items[0].metadata.name}')"
     oc exec "statefulset/${ELASTIC_SET}" -- curl -XDELETE -k --fail "http://127.0.0.1:19200/1000-1000-v1-templates"
     

Reference information

The following information is referenced in this task.