Re: What is this? "Can't find record in '<table>'' on query."

[time e.less]

Adam D wrote:
> Can you still select from the table in question?
> eg select * from tablename limit 1;

Yes. I tried various queries, all of which worked. The most "drastic" was:

mysql> select * from <tablename> into outfile '/tmp/testout';
Query OK, 141506 rows affected (4.98 sec)

I even tried it on the other table in the query, just in case, and that one got the same (non-error) result. I also eyeballed the /tmp/testout file, and the first 8% or so looked fine (although I flipped through it rather quickly and could've missed something).

Just for fun, I did "set global sq_slave_skip_counter = 1" and slave start. Then replication immediately died with this error:

Last_Errno: 1032
Last_Error: Error 'Can't find record in '<tablename>'' on query. Default database: 'production'. Query: 'update <tabname> set counter = counter+1 where userid = 111902'

So then I did a "select * from <tablename> where userid=111902" and I got back a full row of data, which appears to have no corruption or problems whatsoever.

Although, looking at <tablename> here, there are 43 columns, many of which are varchar(255) or varchar(100), and all but two of them are NULLable (the non-NULLable ones are the two PK columns, which are both int(11)). I wonder if I'm running into some strange MySQL cluster bug regarding long rows.

--
timeless