Question RE debugging conflicts in complex transactions

[dvstans]

I've recently switched to RocksDB on ArangoDB 3.6.1 on Ubuntu 18 and I'm now encountering write conflicts I don't understand in some of my transactions. Couple of questions:

1. Is there a way to trace or debug the source of the conflict (document ID or key)?

2. In my test case, there is only one transaction running at a given time, yet I'm still seeing conflict errors. These transactions do update the same document multiple times as they proceed, but it doesn't seem consistent (some sequential updates work, while others do not). Is it possible that back-to-back transactions that update the same documents could conflict due to the updates not being written to disk?

Thanks!

[dvstans]

Update: I did figure out the root cause of my issue. I hadn't read (or forgot) that if a selector for an update call contains a _rev field, and it doesn't match the current rev, a conflict error will be thrown (and no amount of retrying helps!). My mistake was not loading the new version of the document I'd updated previously in the transaction, so the _rev fields were different. (I just change my code to use _id as the selector instead). Question #1 is still relevant.