Fatal lock manager error "Invalid lock id (NNN)" while working with monitoring tables on a highly loaded system [CORE2993] #3375
Submitted by: @dyemanov
Is duplicated by CORE2630
Server crash or freeze are possible as after-effects, and possibly index corruptions as well. It's reproducible easier with v2.5 than with v2.1.
This issue is caused by some logic errors in synchronization related to the monitoring activities. It may occur under quite rare condition when the lock requester gets its AST delivered while it's still waiting inside the lock manager (lock is granted but the thread still cannot check-in back into the engine) and it happens twice subsequently.
The text was updated successfully, but these errors were encountered: