Oracle Database Discussions

Yesterday I made a dump with expdp for 6 instances on my database (dumpfile has a size of 15G after completion) and did this on a oracle 10g RAC database with 2 clusters on the second cluster. Eventually half way through near the end of the dump (which took about 1 hour to complete) my webapplication couldn't connect properly to the database anymore (I run a nagios monitoring tool that complained he couldn't login to the webserver anymore).

I run a web application that connects to these database instances with a connection pool of 20 connections per instance per webserver (4 webservers in total). The load balancing of these connections is handled server side (by the database) and the failover client side (by the webservers). This is what my jdbc connection string looks like:

jdbc:oracle:oci8:@
(DESCRIPTION=
(ADDRESS_LIST=
(ADDRESS=(PROTOCOL=TCP)(HOST=192.168.xxx.xxx)(PORT=1521))
(ADDRESS=(PROTOCOL=TCP)(HOST=192.168.xxx.xxx)(PORT=1521))
)
(LOAD_BALANCE=no)(FAILOVER=yes)
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = xxxxx)
(FAILOVER_MODE =
(TYPE = SELECT)
(METHOD = BASIC)
(RETRIES = 180)
(DELAY = 5)
)
)
)

This is what I executed:
expdp schemas=schema1,schema2,schema3,schema4,schema5,schema6 dumpfile=dump20080204.dmp

I think this happens because the connections either don't get load balanced properly or because the expdp process puts locks on certain tables. Does someone have experience or can see what I'm doing wrong?