Insert of data getting missed/duplicated while using Cursor

Question

Hi all,
Greeting !!
I got a strange issue with cursor usage in a SP . Below is the cursor detail.
Basicially the SP pulls around 2,25,000 records to a temp table(#TMP_ADJ) from there cursor picks and inserts to a physical table(B_SHEET) with some conditions.
Recently we are facing issues in records getting inserted to B_SHEET , the strange part is if this SP is called in a specific period of time i.e every day this job runs at 2 PM GMT and we encountered some of the records getting missed to insert in to B_SHEET
table and some time the record is getting duplicated. Around 200-300 records are getting missed.
we retriggered i.e ran the SP again but this time it was fine, so we confrim that there is no issue with data.
Later we postponed the same job to get triggered at 3 PM GMT post which we dint encounter the problem.
Tough this seems to be not a problem with cursor logic, but i need experts help to identify the exact issue what happens at when the job is triggered at 2 PM GMT.
There has been no change recently done on the SP and the same SP was running fine for past 3-4years.
1 . What are the checks that needs to be done that affects CURSOR.
 DBA has confirmed that CPU utilization / temp db memory are fine during the period of time. And no issues with Data since the second run of the same SP is fine.
Thanks.

Answer 1

If you want suggestions out of the blue, the only recommendation I can give is that you should make sure that the cursor declartion reads:
DECLARE cur CURSOR STATIC LOCAL FOR
The default cursor type is dynamic. Dynamic cursors are difficult to understand and rarely perform well.
Apart from this specific advice, I can only give generic advice since you don't share the code.
It is a common mistake to believe that just because a piece of code produces the expected result that the code is correct. However, the code may only work as intended because of happenstance. For instance a query may be non-deterministic and produces different
results because of the query plan.
Other reasons for "works most of the time, but fails some times" is poor error handling, where you fail to consider errors or handle them inappropriately. In a cursor loop, a common error is also to fail to initiate local variables in the beginning
of the loop, so that they retain their values from the previous iteration.
What I like to emphasise once more is that just because the code works most of the time does not mean that it is correct. The fact that it under some circumstances produces incorrect results, carries much more weight. There is at least one flaw in the code
- you only need to spot it.
Erland Sommarskog, SQL Server MVP, [email protected]