Reversing history: 0 -> 0 [...] Data loss is possible. Aborting.


Searching for this error message on the Internet shows some some results, although nothing seems to apply to my case. In any case apologies if I’m doing something really dumb.

I have a 3 node cluster running mysql+percona just fine (under docker containers). Then I do one of the following:

Case 1:

I (gracefully) terminate one of the nodes and then start it again.

Case 2:

I (gracefully) terminate one of the nodes, delete the grastate.dat and gvwstate.dat and then bring the node up again.

In case 1 I get:

Reversing history: 0 -> 0, this member has applied … more events than the primary component.Data loss is possible. Aborting.

and mysqld terminates. In case 2 I have no error messages and everything seems to work, but obviously this is not the way I should be doing it.

The logs are:
case1 (bad):
case2 (good):

I would be very grateful if someone could give me any hint on what I could be doing wrong.