Oracle.ManagedDataAccess.Core BulkCopy Error

NatLCal

I am encountering the following error when I call WriteToServer and I am not understanding why.

Oracle.ManagedDataAccess.Client.OracleException: 'ORA-01013: user requested cancel of current operation'

Here is the stacktrace:

  at OracleInternal.ServiceObjects.OracleConnectionImpl.CheckForAnyErrorFromDB(TTCError ttcOER)

  at OracleInternal.TTC.TTCDirectPathPrepare.ReadResponse(DirectPathContext& directPathContext)

  at OracleInternal.ServiceObjects.OracleBulkCopyImpl.Prepare()

  at Oracle.ManagedDataAccess.Client.OracleBulkCopy.WriteDataSourceToServer(IBulkCopyDataSource dataSource)

  at Oracle.ManagedDataAccess.Client.OracleBulkCopy.WriteToServer(IDataReader reader)

Alex Keh-Oracle

An ORA-0103 can be indicative that the command is canceled. This can occur if the DB or user initiates an explicit command cancel or the command has a timeout that expired.

NatLCal

The timeout is set to 300, but the error is thrown immediately.

I can't follow the thread past when BulkCopy calls GetOrdinal() in my IDataReader class for each mapped column before the error is thrown.

Does DestinationPartitionName need to be set? I didn't set that property when I was using the older version of Oracle.DataAccess.

Alex Keh-Oracle

That is unusual that the command cancel is immediately initiated. Can you turn on tracing? It would tell us where internally the command was being cancelled. You can add the following two lines of code before opening the connection to enable tracing:

OracleConfiguration.TraceFileLocation = @"<Directory Location>";

OracleConfiguration.TraceLevel = 7;

You can either post the trace or email it to us dotnet_us(at)oracle.com.

NatLCal

MISCELLANEOUS.EXE_PID_22376_DATE_2021_05_21_TIME_10_53_08_539498.txt

Here is the trace. Hope this helps.

NatLCal

@Alex Keh-Oracle, have you had a chance to review the trace file?

Alex Keh-Oracle

I and a dev team member reviewed the trace, but didn't see anything that helped us to understand why the cancel operation occurred.

My recommendation is to open an Oracle Support service request and collect the server side trace. The server side is issuing the cancel call for some reason. It's possible something from ODP.NET is triggering the DB to do that. The server side trace could identify what it is.

Alex Keh-Oracle

@NatLCal I had another team member review the trace that has more expertise with Bulk Copy internals. It appears you are using the same table to bulk copy from and into, EES_RUN_DATA. Can you try changing the destination table name to another table and see if that works? If so, then the bug likely has to do with using the same table name.

Can you also help me understand the use case for bulk copying into the table that already has the data?

NatLCal

@Alex Keh-Oracle, I am using an IDataReader object as the data source. The data is being streamed from a file. I am uncertain how to set a source table name in this case. The Bulk Copy internals must be setting the source table name to the same value as the destination.

Alex Keh-Oracle

Try setting the destination table name to something different, such as:

 blkCpy.DestinationTableName = "EES_RUN_DATA1";

NatLCal

Looks like server tracing is the next step.

Is the ManagedDataAccess client dependent on the server running a specific version (or newer)?