From Elasticsearch/Cloud Search Service
Sample JSON File
"from-config-values": { "configs": [ { "inputs": [ { "name": "fromJobConfig.index", "value": "cdm" }, { "name": "fromJobConfig.type", "value": "es" }, { "name": "fromJobConfig.columnList", "value": "a1:numeric&s1:string" }, { "name": "fromJobConfig.splitNestedField", "value": "true" }, { "name": "fromJobConfig.queryString", "value": "last_name:Smith" } ], "name": "fromJobConfig" } ] }
Parameter Description
Parameter |
Mandatory |
Type |
Description |
---|---|---|---|
fromJobConfig.index |
Yes |
String |
Index of the extracted data, which is similar to the database name in the relational database |
fromJobConfig.type |
Yes |
String |
Type of the extracted data, which is similar to the table name in the relational database |
fromJobConfig.columnList |
No |
String |
List of fields to be extracted. Use & to separate field names. For example, id&gid&name. |
fromJobConfig.splitNestedField |
No |
Boolean |
Whether to split the JSON content of the nested field. For example, a:{ b:{ c:1, d:{ e:2, f:3 } } } can be split into a.b.c, a.b.d.e, and a.b.d.f. |
fromJobConfig.queryString |
No |
String |
Whether to use the Elasticsearch query string to filter the source data. CDM migrates only the data that meets the filter criteria. |
Feedback
Was this page helpful?
Provide feedbackThank you very much for your feedback. We will continue working to improve the documentation.See the reply and handling status in My Cloud VOC.
For any further questions, feel free to contact us through the chatbot.
Chatbot