Details
-
Type:
Task
-
Status: Closed
-
Resolution: Fixed
-
Affects Version/s: None
-
Fix Version/s: None
-
Labels:None
-
# Replies:4
-
Last comment by Customer:false
-
External issue ID:
Description
There is a bug I can't see and missed in the review.
Original comment:
Avoid a deadlock between the sweep server and a checkpoint.
The sweep server was holding a handle exclusive and waiting on the handle list lock (in order to remove the handle from the list), a checkpoint was holding the handle list lock waiting for a read lock on the handle. Found by examining a hang during automated testing.
Issue Links
Activity
autobuild: retest this please
Closing: I prefer the approach in WT-1819.
If we go with the change to sweep in
WT-1812, where we only discard clean files, maybe we can afford to hold the data handle lock around the whole operation, so we don't have this lock ordering problem?