Originally posted by Signalsgt
Originally posted by
Originally posted by RandomBeast
any explanation about why the same WR has his packet dropped/corrupted every sim? If it were an error in the database itself, sure, but if it is a transmission thing...?
Originally posted by Deathblade
Pretty much...we aren't entirely sure if it will even change anything, but it might, so it wouldn't be fair to not add a changelog note.
I still don't buy corrupted packets, TCP traffic simply does not work that way.
Mr. Packet leaves Borts machine with a header, he shows up at the server and the header presents a SYN. The server sends a SYN-ACK handshake back to Borts machine and Mr. Packet along with his friends deliver the data in sequential order. Once all packets have left Bort's machine it sends a FIN. The Server makes sure that all packets have arrived in order and un-corrupted and responds FIN-ACK and shuts down the connection.
Now if Mr. Packet is out of sequence or the data doesn't match the header then the packet is resent, to include a dropped packet.
http://en.wikipedia.org/wiki/File:Tcp_state_diagram_fixed.svg =>Follow the arrows folks.
I have skimmed the last XX pages, and I have more than my share of Perl + MySQL (InnoDB, MyISAM, whatever) troubleshooting under my belt. Bort said somewhere back there that the table in question is so super-huge that a consistency check on it would take an unacceptable amount of time. And he said that MySQL is just saying "goodbye" mid-response.
And it always affects the same players.
I don't see how there's any way this can be anything but a corrupted table. If it were a networking problem, TCP would retransmit, and you should see a delay, but not any change in your query response. If TCP didn't correct the problem, you'd still not see the consistently wrong results -- it'd affect a more random set of players. If it were a bad RAM situation, you'd very likely see more random results (and probably software crashing here or there, too). Disk errors would be equivalent of either corrupt database/table or bad RAM in end-user experience, I'd think.
But, if it's a corrupt table, I don't see how a web server querying the same DB would not see the same symptoms. But stranger things have happened, I'm sure.
There are a few options to fixing broken tables, or at least finding out what's wrong.
The easiest way (to me) is to setup a slave MySQL server, run a mysqldump on the master with the master data intact, import that into the slave, and setup slave replication so that it'll catch up. That'll get a copy of the data as mysqldump sees it, but not the same on-disk image. Unfortunately, that mysqldump can take a damn long time, too. With a lot of careful work, it can be done one table at a time.
Or you can just bite the downtime and run the db check and get it fixed. And hope that's the problem.