if I kill -9 cockroach instance during a SQL DELETE, sometimes I get the errors “read: connection reset by peer” or “connect: connection refused” but the row is deleted anyway !
I handle the error to decide if retry the statement after some seconds or not, but in this scenario I’m not sure if the statement is executed or not.
I got the issue using tx, too. If instance crashes while client tx.commit it seams coach commits the transaction but the client code gets a connection error, then the code handle the error, but the statement is committed !
Any idea how to handle this issue ?