You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The monitoring code unconditionally fetches the header page to report the actual database state. But, if being done at the AST level, this may cause two concurrent pending requests being linked to a single owner. The lock manager highly dislikes this situation and terminates the process via a bugcheck.
It can be reproduced under high load with the SuperClassic architecture. So far I was unable to see this issue on v2.1 Classic, maybe it could be explained by different synchronization techniques used. But anyway, I mark v2.1.0 as possibly affected and the fix will be committed into both branches.
Submitted by: @dyemanov
The monitoring code unconditionally fetches the header page to report the actual database state. But, if being done at the AST level, this may cause two concurrent pending requests being linked to a single owner. The lock manager highly dislikes this situation and terminates the process via a bugcheck.
It can be reproduced under high load with the SuperClassic architecture. So far I was unable to see this issue on v2.1 Classic, maybe it could be explained by different synchronization techniques used. But anyway, I mark v2.1.0 as possibly affected and the fix will be committed into both branches.
Commits: ce7baa7 d55e7eb
The text was updated successfully, but these errors were encountered: