Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Table of Contents

Issues and Decisions

DateIssue/QuestionsDecisions

 

Should dry-run be a path parameter or a query parameter? 

dry-run flag should be an optional query parameter, making it part of path parameter will result in change in multiple endpoints using this feature. List of these APIs can be found below. And changing the path/endpoint by using dry-run as path parameter may impact backwards compatibility of these APIs

Overview

Currently when persisting any data in CPS the data validation is performed using the yang schema and upon successful validation the data is stored in CPS DB. But in case the data validation fails a 500 response is returned by CPS.

...

This is not very helpful for the user as it does not provide proper information about the root cause of failure.

Proposal

The idea here is to have a mechanism to validate the data and throw an appropriate response with message, before without the data is being persisted in the database. This validation check is not meant to be limited for to new data but should also extend to sub-parts sets of a larger data set, such as data used in case of an update operation. So based on this the following approaches are proposed to perform a validation on data:

  • A new endpoint to validate the JSON payload against an existing schema in CPS DB, this endpoint will only validate the data and never persist it.
  • A dry-run flag in the existing CPS API's which would validate the data and return a response containing the details of validation failure. This flag can be implemented to multiple CPS APIs such as POST (Create a Node), PATCH (Update a Node) and PUT (Replace a node), where we can validate the entire data or parts of data and return the exact point of failure in the validation as part of response.

Implementation Ideas

In either case the endpoint should return a REST RESTful response with appropriate response code and message. 

  • Response Code: 400
  • Response Body: a JSON response providing details about the failure. (response body format to be finalized)

List of impacted APIs if dry-run flag is introduced


NameEndpointIssueRequired Action
1Create a Node/{apiVersion}/dataspaces/{dataspace-name}/anchors/{anchor-name}/nodesReturns 500 status on data validation failure

Return 400 status with updated and detailed error message.

Updated Message:

Code Block
{
"status": "400 BAD_REQUEST",
"message": "Data Validation Failed",
"Schema node with name "Node" was not found 
under (urn:ietf:params:xml:ns:netconf:base:1.0)data." 
}


2Update Node Leaves/{apiVersion}/dataspaces/{dataspace-name}/anchors/{anchor-name}/nodesReturns 400 status with entire JSON payload in error message
3Replace Nodes/{apiVersion}/dataspaces/{dataspace-name}/anchors/{anchor-name}/nodesReturns 400 status with entire JSON payload in error message
4Add List Element/{apiVersion}/dataspaces/{dataspace-name}/anchors/{anchor-name}/list-nodesReturns 500 status on data validation failure
5Replace List element/{apiVersion}/dataspaces/{dataspace-name}/anchors/{anchor-name}/list-nodesReturns 400 status with entire JSON payload in error message

Example

The following example shows the expected changes to existing APIs to implement a dry-run flag:

#API

Request parameters

Response CodesScenarios
1

Create a Node:

/v2/dataspaces/{dataspace-name}/anchors/{anchor-name}/nodes


Parameter NameInRequiredDescription
dataspace-namePathtruedataspace name
anchor-namePathtrueanchor name
xpathQuerytruexpath of node
dry-runQueryfalse

flag to enable or disable data validation

(set to false by default)


  • 200 (OK)
    • success
  • 400
    • dataspace not found
      DataspaceNotFoundException
    • anchor not found
      AnchorNotFoundException
    • Data node not found
      DataNodeNotFoundException
    • invalid xpath
      CpsPathException
  • 500
    • unexpected error

Scenario 1: dry run set to false

  • In this case there will be no functional change in the APIs, and they would return appropriate response codes with messages in the following scenarios:
  • In case of successful validation, the data will be persisted, and a 201-response code is returned with nothing in response body
  • In case of validation failure, the data validation error is returned with appropriate message and a 400-response code.

Scenario 2: dry run set to true

  • In this case the APIs will return the same response codes. The only difference will be:
    • in case data validation passes, then it would return a 200-response code without any response body, and the data will not be persisted in case of POST operation or modified in case of PUT/PATCH operations.
    • In case of validation failure, the data validation error is returned with appropriate message and a 400-response code.

Sub-tasks

The following APIs:

  • Create a Node
  • Add List Element

would return a 500-response code in case of Data Validation failure. This is in clear contradiction to the details mentioned in CPS Exceptions and REST APIs HTTP Response Codes documentation. This has been identified as a bug and was fixed in the following patch Fix "Create a node" and "Add List Elements" APIs response code · Gerrit Code Review.

References: