Connector full import synchronization aborts encompassing transaction against large number of deletes (> 500,000 entities)
Tony Sheehy 12 years ago • updated by anonymous 7 years ago • 4
A connector full import/sync that results in a large number of deletions (> 500,000) will abort its respective encompassing TransactionScope.
The following was additionally observed when the TransactionScope was observed to abort:
- When debugging, the TransactionScope itself is registered as complete. (TransactionScope::complete = true)
- From this, it can be potentially inferred that the problem itself happens either on the MSDTC or SQL Server end - as far as the IdB application is concerned, the TransactionScope completed.
- The exception was thrown during IDB ChangeDetection.
- If debugging, ensure that break on all exceptions is enabled.
- During the import, in the respective SQL traces, a substantial number of entries were logged with the following EventClass and ApplicationName.
EventClass: Lock:Timeout ApplicationName: .Net SqlClient Data Provider
Customer support service by UserEcho
The TransactionScope was timing out because all deletions would be pulled in at once (~500,000 entities in test import). This was compounded by the fact that although the TransactionScope timeout was equal to ~2.7 hours, the Maximum available Transaction timeout defaults to 10minutes and is only configurable in the respective host machine.config. This configuration needs to be set in the following way:
Additionally the following can be set to enable editing from the application side:
This however is only a workaround, and the entire Full import paging mechanism will need to be replaced such that entities are paged/deleted in separate transactions.
Reassigned for confirmation of completeness.
Covered by 6.0 Performance in the regression test document