sanity check on innodb config

[kevin foote]

I have two machines in a master <-> master replication scenario. I'm having trouble with start and stop of this thing ever since moving to x86_64 and mysql-ga-5.1.41 .. it takes about 10 minutes to stop then start the mysql process.. the only change I really made during the move was to up the innodb_buffer_pool_size...

host:
os: RHEL5 x86_64
memory: 8G
swap: 4G

I have not adjusted any of the stock hugepage, sysctl, settings my aim was to keep it simple and give the buffer pool more..

Here is my my.cnf pertaining to system / memory type stuff..

-- my.cnf -- clip
max_allowed_packet = 32M
default-character-set = utf8
default-collation = utf8_unicode_ci
log-error = /var/log/mysqld.log
max_connections = 512
log_warnings = 2
query_cache_size = 128M
connect_timeout = 60
max_connect_errors = 1844674407370954751

## INNODB SETTINGS
innodb_data_home_dir = /var/lib/mysql/ibdata
innodb_data_file_path = ib_data1:100M:autoextend
innodb_flush_log_at_trx_commit = 1
innodb_log_group_home_dir = /var/lib/mysql/ibdata_logs
innodb_log_file_size = 256M
innodb_log_files_in_group = 2
innodb_log_buffer_size = 128M
innodb_buffer_pool_size = 4000M
innodb_additional_mem_pool_size = 20M
innodb_support_xa = 1
#innodb_file_per_table = 1
#innodb_flush_method = O_DIRECT

I have not implemented the O_DIRECT or file_per_table yet.. O_DIRECT will be first on my list..

Can anyone point out if the rest of these settings are sane for a box w/8g etc..
and if there is any explination for the slow start/shutdown..

By my calculations mysql should take about ~ 4.5g of memory which should leave plenty for other things.. I have noticed a bit of swapping that is why I want to move to O_DIRECT

any pointers are welcome..

Thanks kevin