Slave with parallel threads is overly optimistic about its master log position after FLUSH LOGS

Description

The test case runs a number ($n) of INSERTs on master, then flushes logs, and tries to synchronize the slave with the master.
If it's executed with slave_parallel_threads > 0, we can see in the slave status that the exec master position is the same as on the master right away, although the number of inserted values differs.
After a while, the position is still the same, but the count increases (up to the expected value, eventually).

The test case shows the problem reliably for me with $n = 100 (the first count on the slave is 20 or so), but if it does not for you, please try to increase $n, I suppose it should increase the probability.

Output:

Environment

None

Assignee

Kristian Nielsen

Reporter

Elena Stepanova

Labels

None

Fix versions

Priority

Major
Configure