How to resolve deadlocks besides using retry?

Question

We have a case to load 100 documents to the BDBXml containers, and we plan to use multi-threads.
Two tests were issued:
1) The 100 documents will be loaded to one container.
5 threads were spawned. While in one thread, the document was being loaded, in the other threads, the loading operations were blocked(DB_LOCK_DEADLOCK exception was thrown). So we introduce the retry strategy. Basically we did 20 times retry every 0.25 seconds.
Unfortunately, even we used up all the retries, there were still many loading operation failed.
2)The 100 document will be loaded to 100 containers.
5 threads were spawned. We didn't meet the deadlock issue. All the loading operation succeeded.
We prefer to using one container for all the 100 documents, but we could not resolve the deadlock issue even we use the retry.
Is there any other way to resolve this issue?

Answer 1

Yes, I was talking about loading all documents into one container. I didn't completely get your last question, but it's up to developer whether he wants to retry the deadlocked operation or just skip it. In your case you want of course to retry since you want to load all documents. I'm not sure that it should be faster if you load documents from multiple threads given that deadlocks are unavoidable. Did you try to switch auto-indexing off and see whether it will be faster to load all the documents sequentially or whether it eliminates some deadlocks?
Vyacheslav