a) Has it been always 1032 or has it been 1062 (Duplicate entry) anytime?
b) What is the sysbench that is running (since there is a Delete_rows event). Also, I presume, from description, that sysbench is running while you SIGKILL the other node's mysqld?
c) Now, from ecec24ed-ab72-11e2-0800-008b24f29e8e:194044 and the error, it looks like it is failing at the very next event.
Now, this means, when it got killed, only the events upto a seqno lower than 194044 actually made to InnoDB whereas to the log it was recorded as 194044.
What is the value of innodb-flush-log-at-trx-commit.
@Alexey,
How do the values flushed to log interact with innodb-flush-log-at-trx-commit.
Interesting.
a) Has it been always 1032 or has it been 1062 (Duplicate entry) anytime?
b) What is the sysbench that is running (since there is a Delete_rows event). Also, I presume, from description, that sysbench is running while you SIGKILL the other node's mysqld?
c) Now, from ecec24ed- ab72-11e2- 0800-008b24f29e 8e:194044 and the error, it looks like it is failing at the very next event.
Now, this means, when it got killed, only the events upto a seqno lower than 194044 actually made to InnoDB whereas to the log it was recorded as 194044.
What is the value of innodb- flush-log- at-trx- commit.
@Alexey,
How do the values flushed to log interact with innodb- flush-log- at-trx- commit.