Skip to Content
0

SAP Cloud Platform Predictive Service Recommendations API error - EXX110

May 13, 2017 at 06:24 PM

90

avatar image

Hi,

I registered a dataset and want to estimate the costs of creating a recommendation model using /api/analytics/recommendations/recommender/guess/sync resource.

This is the request body:

{
    "transactionData" : {
        "datasetID" : 9,
        "transaction" : {
            "userColumn" : "CARD_ID",
            "itemColumn" : "PRODUCT_ID"
        }
    },
    "modelingSettings" : {
    "minimumSupport" : 10,
    "minimumConfidence" : 0.5
    }
}

However, when I make my request, I get the following response:

{
  "errors": [
    {
      "errorCode": "EXX110",
      "errorMessage": "An SQL error has occurred."
    }
  ]
}

Of course, I get the same error if I actually try to create a recommendation model.

I have attached the logs that are created after I make requests. What could be the reason for the error that I am getting? As I noticed in the logs, it says there "SAP DBTech JDBC: [2048]: column store error: search table error: [9] Memory allocation failed" . Could it be because HANA table that is registered as the dataset I'm using is too large? It's a row-based table that has only 7 columns and 110000 records.

I also noticed that if I change my request body so that it also indicates a dateColumn and period inputs and if I set a very short period for startDate-endDate, I do get a response. Could the whole problem be because aac4paservices app can't use up too much memory and I need to somehow increase the amount of memory it can use?

Here is the modified body of a request that does bring me proper response.

{
    "transactionData" : {
        "datasetID" : 9,
        "transaction" : {
            "userColumn" : "CARD_ID",
            "itemColumn" : "PRODUCT_ID",
          "dateColumn" : "DATE"
        },
      "period" :{                                  
        "startDate" : "2016-02-01",                      
        "endDate" : "2016-02-02"                        
    }
    },
    "modelingSettings" : {
    "minimumSupport" : 10,
    "minimumConfidence" : 0.5
    }
}

1.txt2.txt3.txt

For reference, I have a trial SAP CP account and I'm using HANA MDC system.

1.txt (270 B)
2.txt (7.2 kB)
3.txt (7.4 kB)
10 |10000 characters needed characters left characters exceeded
* Please Login or Register to Answer, Follow or Comment.

1 Answer

Best Answer
Abdel DADOUCHE
May 16, 2017 at 04:18 PM
0

Hi,

The SAP CP HANA MDC instances have only 6 GB of RAM which means that it has it's limits.

I would first recommend you to use column store instead of row store, as it will save you HANA memory to load that table.

Then, the aac4paservices application is not building the model itself, as everything is happening inside of your HANA instance via the use of the APL (Automated Predictive Library).

Can you check how many distinct card id and product id do you have in your dataset? Also how many rows in your dataset do you have between 2016-02-01 and 2016-02-02?

Share
10 |10000 characters needed characters left characters exceeded