memlock was the issue. turns out, memlock settings were at default (see above), and this might have been preventing the allocation of 4+gb of memory to mysql.

changed memlock settings in '/etc/security/limits.conf', and limit to 8GB. then, allocated 7GB to mysql buffer pool. worked. mysql now starts and stops without throwing any errors in the log.

a few things to note here -

1. from what i have read online, the amount of memory allocated to mysql is slightly more than the setting for buffer pool. so lets say you assign 7GB to pool, actual memory assigned to mysql would be closer to 8gb (this is an example, not real numbers).

2. turning the mysql service on/off has to be done very carefully. i have noticed that when i do a mysqld-start followed by quickly by a mysqld-stop, mysql responds with [FAILED]. logs says allocation of 7GB failed.

BUFFER POOL INSTANCES

The rule of thumb I usually use is based on a special program in in the Linux environment

numactl --hardware

When I run this, I get the following output

sh-4.1# numactl --hardware
available: 4 nodes (0-3)
node 0 cpus: 0 1 2 3
node 0 size: 49151 MB
node 0 free: 241 MB
node 1 cpus: 4 5 6 7
node 1 size: 32768 MB
node 1 free: 39 MB
node 2 cpus: 8 9 10 11
node 2 size: 49152 MB
node 2 free: 49 MB
node 3 cpus: 12 13 14 15
node 3 size: 32752 MB
node 3 free: 32 MB
node distances:
node   0   1   2   3
  0:  10  16  16  16
  1:  16  10  16  16
  2:  16  16  10  16
  3:  16  16  16  10

This quickly tells me how many CPUs and cores I have on my DB Server.

In general, I set the innodb_buffer_pool_instances to the number of physical CPUs or the number of cores. In your particular, I set the innodb_buffer_pool_instances to 4 or 16.

BUFFER POOL SIZE

If you make the InnoDB Buffer Pool bigger than 50% of the installed RAM, mysqld causes the OS to start swapping ... BADLY !!!! If you need a Buffer Pool that big, then tuning innodb_buffer_pool_instances becomes even more critical.

EPILOGUE

As soon as innodb_buffer_pool_instances was first introduced, I immediately experimented with it. I had a client that had 192GB DB Server, dual hexacore with a 162GB Buffer Pool. I simply set it to 2 and everything worked out just fine. Please see my old post from Feb 12, 2011 on this : How do you tune MySQL for a heavy InnoDB workload?

GIVE IT A TRY !!!