Re: Can replicating between versions cause crashed tables?

[Cody Konior]

To answer all your questions:
- Yes Master 4.0 Slave 5.1.
- The 4.0 didn't have a my.ini file, instead it ran mysqld_safe with 3 command-line parameters. I added those to the default 5.1 file.
max_allowed_packet=16M
max_connections=500 (actually increased this to 800)
open_files=2048

- During the up-converion I made the tables latin1. No tables should have odd characters, it's all plain-text and numbers and plain-text XML.

- Yes we're using MyISAM because that's what the original tables were in. I've been reading about it and InnoDB lately - it seems that InnoDB has crash recovery and MyISAM doesn't (?) so I'm not sure if I should switch over when I do the switch over.

I noticed replication had stopped, so I checked all the tables and started it again. It immediately corrupted 1/3 of the tables (the commonly used ones). The reason listed for each table was:

Wrong bytesec: 50-54-0 at linkstart: 3016236
Corrupt
Unexpected byte: 3 at link: 3343644
Corrupt
1 client is using or hasn't closed the table properly
1 client is using or hasn't closed the table properly
Size of datafile is: 350752 Should be: 350712
Record at: 350712 Can't find key for index: 1
Corrupt

That's a lot. I think replication between these versions must just not work well?