MySQL 5.7 not accepting values of innodb_buffer_pool_instances greater than 1

Yes it is safe to delete the log file once mysqld has been shutdown

In light of this, just perform the following steps:

mysql -uroot -p... -e"SET GLOBAL innodb_fast_shutdown = 0"
service mysql stop
mv /var/lib/mysql/ib_logfile[01] /tmp
service mysql start

Starting up mysqld will recreate ib_logfile0 and ib_logfile1

Give it a Try !!!

UPDATE 2011-10-20 16:40 EDT

It cleanly page out all data in the InnoDB Buffer Pool prior to redoing the Log Files, you should set this option about 1 hour before shutdown:

SET GLOBAL innodb_max_dirty_pages_pct = 0;

By default, innodb_max_dirty_pages_pct is 75 (MySQL 5.5+) or 90 (prior to MySQL 5.5). Setting this to zero keeps the number of dirty pages under 1% of the InnoDB Buffer Pool. Performing service mysql stop does this anyway. In addition, a shutdown will finish up any remaining items in the redo log. To keep to this option just add it to /etc/my.cnf:

[mysqld]
innodb_max_dirty_pages_pct = 0

UPDATE 2013-04-19 16:16 EDT

I updated my answer a little more with innodb_fast_shutdown because I used to restart mysql and stop mysql to do this. Now, this one-step is vital because every transaction uncommitted may have other moving parts within and outside of the InnoDB Transaction Logs (See InnoDB Infrastructure).

Please note that setting innodb_fast_shutdown to 2 would clean the logs out as well but more moving parts still exist and gets picked on Crash Recovery during mysqld's startup. Setting of 0 is best.

Amit, the problem is that this variable is set to reload a copy of the buffer pool dumped to disk during a stop/restart process. In order to achieve this you need to set innodb_buffer_pool_dump_at_shutdown to ON prior to restart your server, please check this link for further info: https://dev.mysql.com/doc/refman/5.6/en/innodb-preload-buffer-pool.html Regarding NUMA configuration this has nothing to do with buffer pool fill but how memory dedicated to buffer pool is balanced between CPU nodes (this is useful only for physical servers) so in modern hardware where you have more than one physical CPU you will avoid having swapping issues by balancing large portions of memory allocation (like buffer pool normally has) between nodes. Hope this helps.