I think I found the reason. When XtraBackup reads a log block that has not been completely written by the server yet, it will fail with an error like that. Which is a regression introduced in 2.0. XtraBackup 1.6 will print a warning, but will retry reading the same block later.
I think I found the reason. When XtraBackup reads a log block that has not been completely written by the server yet, it will fail with an error like that. Which is a regression introduced in 2.0. XtraBackup 1.6 will print a warning, but will retry reading the same block later.
I reported this problem as bug #1015416.