Table size difference between master and slaves

[Benny Pei]

Hi all,

I have a table like below. This table went through insert,update,delete. I have a two slaves setup to replicate from a master. The table structure is like this:

CREATE TABLE mytable (
`myid` varchar(20) NOT NULL,
`last_num` bigint(20) NOT NULL,
`sessdata` mediumblob,
PRIMARY KEY (`myid`))

The table contains about 8K rows average and the size is about 800MB. One of the slaves behaves differently and the table size is about 2.8GB. I checked the other slave and the size is about the same as the master. So one day I went ahead and rebuilt the table in the slave to reduce the table size. A few months later, the table size grew again. I couldn't figure out what might be wrong. I see both slaves are using ROW based replication. Could someone tell me what might have caused the difference in size? Both slaves were set as readonly. I am running version 5.5 on Solaris 10.

thanks,

benny