Details
-
Bug
-
Status: Open
-
Normal
-
Resolution: Unresolved
-
None
-
None
-
Normal
Description
It's possible for 2 compactions to be competing for an amount of disk space smaller than either one would take up by itself but enough so that the system runs out of disk.
Cassandra should:
- factor in the estimated disk usage of current running compactions for each drive, and subtract that amount from the space available (do not overcommit)
- detect when several compactions will be competing for space and prioritize ones that can complete given the space available.