IBM Support

Watson Explorer V12.0.2 known issues on oneWEX for ICP

Fix Readme


Abstract

This document describes the know issues found in Watson Explorer Version 12.0.2 on oneWEX for ICP deployment.

Content

Using GlusterFS as Persistent Volume with oneWEX for ICP

When you are using IBM Cloud Private 3.1 with GlusterFS, you might see some storage issues. See Troubleshooting and support on storage for ICP 3.1 for detailed steps of the solution.


Importing classifier/ranker related collection into oneWEX for ICP

When exporting the collection that includes classifier/ranker configuration or classifier/ranker is deployed then importing the collection into oneWEX for ICP, other collections on the imported system may not be able to process indexing. To work around the issue, update the configuration of the imported collection on oneWEX for ICP with REST API as follows:

Note: The configuration cannot be changed in Watson Explorer Admin Console. Use REST API.

1. Get the collection id of all collections which have the the specific properties with following REST APIs on the system that the collection is imported. 

  • exported.server.info property with https://<ONEWEX_SERVER>/api/v1/collections
  • mlCollectionId property with https://<ONEWEX_SERVER>/api/v1/sdrankers
  • SimilarDocumentRanker property with https://<ONEWEX_SERVER>/api/v1/labelers

 

curl -X GET 'https://<ONEWEX_SERVER>/api/v1/collections' -H 'Content-Type: application/json' -H 'authorization: Basic <Base64 encoded username:password>' -k -o col_list.json
curl -X GET 'https://<ONEWEX_SERVER>/api/v1/sdrankers' -H 'Content-Type: application/json' -H 'authorization: Basic <Base64 encoded username:password>' -k -o ranker_list.json
curl -X GET 'https://<ONEWEX_SERVER>/api/v1/labelers' -H 'Content-Type: application/json' -H 'authorization: Basic <Base64 encoded username:password>' -k -o labeler_list.json

 

2. Get the collection configuration of those collections identified in the previous step with the REST API and save it as a file.
For example, save the output to test_col.json file.

curl -X GET 'https://<ONEWEX_SERVER>/api/v1/collections/<COLLECTION_ID> ' -H 'Content-Type: application/json' -H 'authorization: Basic <Base64 encoded username:password>' -k -o test_col.json

 

3. Add "compaction.enabled": false in tags property in the saved json file.

{
    "id": "3dbf7d7d-b9d2-1180-0000-0166b9056b9a",
    "name": "test",
...
    },
    "tags": {
        "compaction.enabled": false,
        "documentLevelSecurity": "none",
...

}

 

4. Update the collection configuration with the REST API using updated json file.

curl -X PUT 'https://<ONEWEX_SERVER>/api/v1/collections/<COLLECTION_ID> ' -H 'Content-Type: application/json' -H 'authorization: Basic <Base64 encoded username:password>' -k --d @test_col.json

 


Using Character Pattern annotator on oneWEX for ICP

When oneWEX for ICP is deployed with default Helm chart parameters, Character Pattern annotator will show the different facet counting result compared to the result on oneWEX for single container deployment. To work around the issue, update the following Helm chart parameters and deploy the oneWEX for ICP.

  1. Set "Number of replicas of saved data" parameter as 1 for HDP service. (default value: 2)
  2. Set "Available memory of HDP worker process" parameter as 8192 (or larger) for HDP service. (default value: 4096)

 

See Helm chart parameters for Watson Explorer oneWEX on IBM Cloud Private (version 12.0.2 and later) for detailed information about each parameter.


Utilizing facet requests on UI with oneWEX for ICP

When oneWEX for ICP is deployed with multiple Discovery service replicas, you may see following issues on oneWEX UI:

  1. Facet counting on Content Miner UI will return no result.
  2. Query Assistance features used in following UI will not work as expected, such as:
    1. On Content Miner UI, "Analyze cause or characteristics" will not return the result.
    2. Using Built-in Query Recommendation widget with "Words or phrases correlated to the input" type on oneWEX Application Builder will not return the result.

Note: Default Helm chart is configured to use Discovery service replicas as 1. When oneWEX for ICP is deployed with default Helm configuration, this issue will not occur. See Helm chart parameters for Watson Explorer oneWEX on IBM Cloud Private (version 12.0.2 and later) for detailed information about each parameter.

To work around the issue, update the configuration with the REST API as follows:

1. Identify the collection id that you are using, and get the configuration of the collection with the REST API, and save it as a file.
For example, save the output to params.json file.

curl -X GET 'https://<ONEWEX_SERVER>/api/v1/collections/<COLLECTION_ID>/exploration/configparams' -H 'Content-Type: application/json' -H 'authorization: Basic <Base64 encoded username:password>' -k -o params.json

2. Add "facet.overrequest.ratio":3 to in the saved json file. For example,

            "df": "_text_",
            "qf": "_text_",
            "carrot.title": "id",
            "SecurityFilteringRequest": "{\"documentLevelSecurity\":\"none\"}",
            "facet.overrequest.ratio":3

3. Update the collection configuration with the REST API using updated json file.

curl -X POST  'https://<ONEWEX_SERVER>/api/v1/collections/<COLLECTION_ID>/exploration/configparams' --d @params.json -H 'Content-type: application/json' -k

Note: The facet counting issue can be worked around by updating the collection configuration as above. However, query assistance features may not be able to be worked around depending on the data volume that you process.

 

[{"Business Unit":{"code":"BU053","label":"Cloud & Data Platform"},"Product":{"code":"SS8NLW","label":"Watson Explorer"},"Component":"oneWEX;","Platform":[{"code":"PF016","label":"Linux"}],"Version":"12.0.2","Edition":"","Line of Business":{"code":"LOB10","label":"Data and AI"}}]

Document Information

Modified date:
02 November 2018

UID

ibm10738207