Hi All,
An update to a previous post, the history is at the bottom. To summarise:
1. Identical package running on two machines (DevServer and LiveServer) every day.
2. On DevServer machine it never fails, on LiveServer it fails occasionally.
3. Every day where there is no error the rowcounts on both machines are exactly matched. Package details
On days when it errors
1. Stuff happens
2. Final step is outputting into Table1 (a well formated SQL Server Table (int, nvarchar, datetime))
3. Redirect error rows into Table2 (same as Table 1 but all fields nvarchar(255))
Today Table1 on LiveServer was 200+ rows short of Table1 on DevServer. You would expect Table2 to contain 200+ rows right. Nope, it contains 40!! An there is loads of empty strings in various fields, no consistency to which fields are mssing data. I have looked at the rows on DevServer which are missing on LiveServer but there is nothing obviously strange about them (no $%^&*()_ or anything)
Now I am stumped. Clearly records are being lost, but some are vanishing into the ether. I can give more detail of the upstream tasks in this package, I just didn't want to muddy the waters.
N.B. I reran the package 10minutes later after making NO CHANGES AT ALL and it ran perfectly
Help, I'm going mad now!!!
David
---------------------Previously on "ghost" row!!--------------------------------------------------
I have an SSIS package which fails periodically:
"There was an error with input column "uploaddateintl" (68541) on input "OLE DB Destination Input" (68526). The column status returned was: "Conversion failed because the data value overflowed the specified type."
Sounds like a simple data error right!! Mmmm......
-- I have rerun the package (takes ~ 5mins ) to discount any data errors (i.e. same source file which is a CSV).
-- It runs 99% of the time sometimes we have to run it repetitively all morning just to get it to error.
-- In the live environment it fails over 2-3 times per week, we have scheduled it to restart if it errors and it re-runs fine
-- Maybe once this month it has fallen over a second time, but re-run perfectly on third attempt.
Today tried a new tack and output the erroring rows into a table. This didn't work as the conversion error still exists into the error table (obviously Doh!!) So I changed the date fields to nvarchar(50). And it output an empty row into the table. If the input to the Destination has empty Date fields I'm not surprised it errors, but the question is where is the empty row coming from and why does it only appear some of the time when the same source data is being used?? I ran this with the same file over and over and it only failed once in a whole morning...........
-------------------------------------------------------------------------------------------------------