Oracle Database Inserts Via Microsoft Data Transformation Services (DTS)

Question

This question involves a SQL Server database and an Oracle database. The databases reside on different servers. One of our developers periodically uses Microsoft DTS (Data Transformation Services) to read data from a SQL Server database and insert it into an Oracle database. Normally the job runs once a day and reportedly inserts about 20,000 rows. The job usually runs fine. About a month ago execution of the daily job was suspended. Two days ago the developer ran a job to select and insert nine days of information. He estimated that 80,000 rows would be inserted. The job cancelled after twenty-three minutes when it filled up the 512 MB UNDO tablespace. (FYI, we use automatic UNDO management.) At the point of failure the number of active sessions spiked sharply in the Oracle database because of system I/O waits (log file parallel write, db file parallel write, and control file parallel write). The number of active sessions also spiked sharply in three other Oracle databases whose files reside on the same array of disk drives. Most of those sessions were waiting on commits (log file sync). The spikes lasted for one minute or less. Grid Control’s performance monitor shows that sqlservr.exe is the module being executed when the UNDO tablespace fills up. We ran the job a second time and closely monitored it, watching the amount of UNDO space grow until it used all 512 MB available. The symptoms described above for the first cancellation were repeated in the second cancellation.
We reran the job by processing a single day’s worth of information and that ran fine. Then we ran it for two days of information, then for six days of information. Everything ran fine. During those tests no more than 70 MB of space of UNDO were used.
Our developer reported that last week he ran the job for nine days of information, the same amount as the job that cancelled twice today. He estimates that it ran for about 80 minutes and went to a normal end-of-job.
Can anyone here offer an explanation of why we seem to be getting these varied demands for space in the UNDO tablespace? Do you know if Microsoft DTS issues a commit after each insert or only a single commit at the end-of-job?
Thank you,
Bill

Answer 1

Hi Arthur,
Yes both instances are same.
Microsoft SQL Server 2008 R2 (SP2) - 10.50.4263.0 (X64)   Aug 23 2012 15:56:56   Copyright (c) Microsoft Corporation  Enterprise Edition (64-bit) on Windows NT 6.1 <X64> (Build 7601: Service Pack 1) (Hypervisor)
I have run the Main package using the using SQL Agent, main package calls the child packages.
The error message shown on SQL agent job is:
R6025  - pure virtual function call.  The return value was unknown.  The process exit code was 255.
or sometimes 
The step did not generate any output.  The return value was unknown.  The process exit code was -532459699.
in the even log it says:
Error Level:
Event ID 1000
Faulting application name: DTExec.exe, version: 2009.100.4263.0, time stamp: 0x5036ba73
Faulting module name: DTSPipeline.dll, version: 2009.100.4263.0, time stamp: 0x5036ba53
Exception code: 0x40000015
Fault offset: 0x00000000000a33c5
Faulting process id: 0x98c
Faulting application start time: 0x01cf64ba9b72b27c
Faulting application path: C:\Program Files\Microsoft SQL Server\100\DTS\Binn\DTExec.exe
Faulting module path: C:\Program Files\Microsoft SQL Server\100\DTS\Binn\DTSPipeline.dll
Report Id: e8eb9b4f-d0ad-11e3-babd-005056997b14
Information Level:
Windows error reporting  Event ID 1001
Fault bucket , type 0
Event Name: APPCRASH
Response: Not available
Cab Id: 0
Problem signature:
P1: DTExec.exe
P2: 2009.100.4263.0
P3: 5036ba73
P4: DTSPipeline.dll
P5: 2009.100.4263.0
P6: 5036ba53
P7: 40000015
P8: 00000000000a33c5
P9:
P10:
Attached files:
These files may be available here:
C:\ProgramData\Microsoft\Windows\WER\ReportQueue\AppCrash_DTExec.exe_ccc7a4e176faafbea69955957371ea96e175b_44c41e3e
Analysis symbol:
Rechecking for solution: 0
Report Id: e8eb9b4f-d0ad-11e3-babd-005056997b14
Report Status: 4