Linux cross-connect routing problem

[Dal]

I have two RedHat EL 4 boxes linked via a cross-connect. One is a web server (10.0.0.3) and one is a mySQL server (10.0.0.2), the interface between them is eth1 on both machines and a second interface eth0 connects to the internet. I want to use the web server to send queries to the database server via eth1, 10.0.0.2:3306 in this case. If I send a database query via eth1 there is a delay of about 10-20 seconds before the result comes back. If I send the same query to the database server but use it's main IP instead of the internal IP so that the query is being sent to it over the internet (xx.xx.xx.xx:3306), the result comes back instantly. Similarly, if I send a query from any remote server the result is instant.

Why should there be such a huge delay when sending a query directly through the cross-connect ?

The routing table ( ip route show ) for the web server is:

xx.xx.xx.xx/xx dev eth0 proto kernel scope link src xx.xxx.xx.xx
10.0.0.0/24 dev eth1 proto kernel scope link src 10.0.0.3
default via xx.xx.xx.xx dev eth0

and the routing table on the database server is:

xx.xx.xx.xx/xx dev eth0 proto kernel scope link src xx.xx.xx.xx
10.0.0.0/8 dev eth1 proto kernel scope link src 10.0.0.2
default via xx.xx.xx.xx dev eth0

I have ifcfg-eth1 on both boxes:

DEVICE=eth1
ONBOOT=yes
TYPE=Ethernet
IPADDR=10.0.0.3 / 10.0.0.2
NETMASK=255.255.255.0

Both boxes can ping each other and transfer files using wget without any apparant problems or delays.

Anyone have any ideas on how to fix this 10-20 second delay when sending queries through the cross-connect ?

Thanks in advance.

[Dal]

Ah finally found out what was causing this. Turns out it had nothing to do with the routing tables at all, the problem seems to be a bug in mySQL involving looking up the hostnames of users connecting to the database.

The solution was to add "skip-name-resolve" to my.cnf.