ORNL
DB - Integrity check misses task lock missing endpoint
Not sure how the problem happened, but there is a task lock that refers to a non-existent task. The DB integrity check should catch this kind of error.
I do not understand the specific nature of this problem. Is this an issue of resolving a failed request? Do we need to have a cleanup service runs that will automatically flag and fix where possible issues like this (network failures, request failures due to limited resources)? Or is this indicating poorly implemented logic (coding errors - edge cases).
The issue is that the DB should never (ever!) lose referential integrity (RI). Some kind of bug caused a vertex to be deleted but left an edge that referred to it - this is a loss of RI. I created a DB admin check method that crawls the entire DB looking for RI issues, but it did not catch this case, so the check method needs to be fixed/updated.