OBSERVATION #1

Since your buffer pool is 6G (6144M), the innodb_log_file_size should be 1536M (25% of 6G)

OBSERVATION #2

You have sync_binlog set to 1. This provides the safest ACID compliant setup. It can also slow things down dramatically. You say it seems as though its writing each row 1 at a time to disk. That's the case because each completed DML (INSERT, UPDATE, DELETE) and DDL (ALTER TABLE) statement gets written to the binary logs. The default for sync_binlog is 0. That let's the OS be responsible for flush binary log changes to disk.

OBSERVATION #3

You have innodb_io_capacity set at 10000. That's really 10000 IOPs you are expecting of mysql. Try lowering it.

There are some things to do in this respect

• When it comes to sync_binlog
  • Comment it out of my.cnf
  • Set sync_binlog to 100 or 200 to pace the flushing of binary logs
• You should make sure data disk has a battery-backed cache (with a good battery)

RECOMMENDATION

STEP 01) Set these options in my.cnf

[mysqld]
innodb_fast_shutdown = 0
innodb_log_file_size = 1536M
sync_binlog = ( you choose 0 or 100 )

STEP 02) Remove innodb_io_capacity. Let it be set to default.

STEP 03) Run this command in mysql

mysql> SET GLOBAL innodb_fast_shutdown = 0;

STEP 04) Apply settings

cd /mnt/Storage/mysql
service mysql stop
mv ib_logfile0 ib_logfile0.bak
mv ib_logfile2 ib_logfile1.bak
service mysql start

This should speed things up

I was having the exact same problem on an AWS t1.micro that has about .6GB of memory. It would go through the entire setup process then fail on initializing mysqld. As it turns out, that instance did not have enough memory to get percona server 5.6 going. I brought up a faster instance running 3.75GB of memory and it started without issue.

If I were to speculate, I would guess you are on the f1-micro instance which is why you are getting the "Cannot allocate memory" errors. Try the g1-small instance and see if that is enough to get mysql running.

Update:
I was/am using Percona's default configuration and I was able to successfully install on a Digital Ocean droplet with 1GB of memory. So, it seems as though the magic number of needed memory is between .6GB and 1GB.