Shantanu's Blog

Create and configure load-balancer:
elb-create-lb MyLB --availability-zones us-east-1b --listener "protocol=TCP, lb-port=3306, instance-port=3306"

elb-configure-healthcheck MyLB --target "TCP:3306" --interval 30 --timeout 2 --healthy-threshold 6 --unhealthy-threshold 2

elb-register-instances-with-lb MyLB --instances i-29184c43

Create config parameter and auto-scaling group:
as-create-launch-config MyLC --image-id ami-41d00528 --instance-type m1.small --key severalnines --group quick-start-1 

as-create-auto-scaling-group MyAutoScalingGroup --launch-configuration MyLC --availability-zones us-east-1a  --min-size 2 --max-size 5 --load-balancers MyLB

Create policies and alarms:
as-put-scaling-policy MyScaleUpPolicy --auto-scaling-group MyAutoScalingGroup  --adjustment=1 --type ChangeInCapacity  --cooldown 300
 
mon-put-metric-alarm MyHighCPUAlarm  --comparison-operator  GreaterThanThreshold  --evaluation-periods  1 --metric-name  CPUUtilization  --namespace  "AWS/EC2"  --period  600  --statistic Average --threshold  80 --alarm-actions Policy_from_previous_step --dimensions "AutoScalingGroupName=MyAutoScalingGroup"

as-put-scaling-policy MyScaleDownPolicy --auto-scaling-group MyAutoScalingGroup  --adjustment=-1 --type ChangeInCapacity  --cooldown 300

mon-put-metric-alarm MyLowCPUAlarm  --comparison-operator  LessThanThreshold --evaluation-periods  1 --metric-name  CPUUtilization --namespace  "AWS/EC2"  --period  600  --statistic Average --threshold 40 --alarm-actions Policy_from_previous_step --dimensions "AutoScalingGroupName=MyAutoScalingGroup"


Describe scaling: 

as-describe-auto-scaling-groups
as-describe-scaling-activities
as-describe-launch-configs

associate domain:
## associate and dis-associate route 53 domain to LB
elb-associate-route53-hosted-zone useast --rr-name mydb.shantanuoak.com --hosted-zone-id Z36XR238KZYP2F --weight 100

elb-disassociate-route53-hosted-zone useast --rr-name mydb.shantanuoak.com --hosted-zone-id Z36XR238KZYP2F --weight 100

Pause auto-scale:

as-suspend-processes MyAutoScalingGroup
as-resume-processes  MyAutoScalingGroup

Delete config parameter and auto scale group:
as-delete-auto-scaling-group MyAutoScalingGroup  --force-delete

as-delete-launch-config  MyLC

Replace launch config:

as-create-launch-config app-server-launch-config-2 --image-id ami-c503e8ac --instance-type c1.medium --group web

as-update-auto-scaling-group app-server-as-group-1 --launch-configuration app-server-launch-config-2

Terminate instances:
as-update-auto-scaling-group app-server-as-group-1 --min-size 0

as-terminate-instance-in-auto-scaling-group i-12345abc --decrement-desired-capacity

as-terminate-instance-in-auto-scaling-group i-12345abc --no-decrement-desired-capacity

Here is the shell script that will install mysql version 5.5 on a new instance.
sh -xv /root/clean_install.sh

#!/bin/sh

# mysql data directory
my_data_dir=/data/mysql/jun19

mydate=`date '+%j%H%M%S'`

## make sure new EBS is formatted
#dmesg
#time mkfs /dev/xvdt

## Add mount point if any and mysql start command in /etc/rc.local

#/bin/mkdir -p /data
#/bin/mount /dev/xvdab /data
#/etc/init.d/iptables stop
#/etc/init.d/mysql start

## disable selinux
sed -i 's/SELINUX=enforcing/SELINUX=disabled/' /etc/sysconfig/selinux
/usr/sbin/setenforce 0
echo 0 > /selinux/enforce

## shut-down mysql if already running
mysqladmin shutdown

# remove old data directory
#rm -rf /var/lib/mysql/
rm -rf /root/download

## create required directories
# datadir
mkdir -p $my_data_dir
# pid directory
mkdir -p /var/run/mysql
# default socket directory
mkdir -p /var/lib/mysql
# download directory
mkdir /root/download
cd /root/download

wget http://files.directadmin.com/services/all/mysql/64-bit/5.5.20/MySQL-client-5.5.20-1.linux2.6.x86_64.rpm
wget http://files.directadmin.com/services/all/mysql/64-bit/5.5.20/MySQL-devel-5.5.20-1.linux2.6.x86_64.rpm
wget http://files.directadmin.com/services/all/mysql/64-bit/5.5.20/MySQL-server-5.5.20-1.linux2.6.x86_64.rpm
wget http://files.directadmin.com/services/all/mysql/64-bit/5.5.20/MySQL-shared-5.5.20-1.linux2.6.x86_64.rpm

# create my.cnf file
cat > /etc/my.cnf << heredoc

[mysqld]
datadir=$my_data_dir
socket=/var/lib/mysql/mysql.sock
user=root

max_connections=2000
max_connect_errors=18446744073709547520
open_files_limit=40000

#slave-skip-errors=1062

## logging
relay-log=slave-relay-bin
log-bin=master-bin

slow_query_log=1
slow_query_log_file=/mnt/slow.log
long_query_time=2
log-slow-slave-statements

## date %j%H%M%Sserver-id=$mydate

## network and slave needs more bandwidth
max_allowed_packet=256M

## temporary tables optimization
tmp_table_size=256M
max_heap_table_size=256M

## join queries use this buffer
read_buffer_size=32M

## Query Cache
query_cache_type = 1
query_cache_size = 200M

## Performance Tunning
key_buffer_size=1024M

## temp dir
tmpdir = /mnt/

## InnoDB optimization
innodb_buffer_pool_size=1024M
innodb_log_file_size=64M
innodb_file_per_table=1
innodb_flush_method=O_DIRECT

[mysqld_multi]
mysqld = /usr/bin/mysqld_safe
mysqladmin = /usr/bin/mysqladmin
user = multi_admin
password = multipass

[mysqld2]
socket = /tmp/mysql.sock2
port = 3307
pid-file = /mnt/data/hostname.pid2
datadir = /mnt/data/

[mysqld_safe]
log-error=/var/log/mysqld.log
pid-file=/var/run/mysql/mysqld.pid

heredoc


# reset data-directory
mydir=`grep ^datadir /etc/my.cnf | awk -F'=' '{print $2}'`
rm -rf $mydir/*

# remove mysql
for package in `rpm -qa | grep -i mysql`
do
rpm -e $package
done

rpm -e mysql-libs-5.1.52-1.el6_0.1.x86_64

# install mysql
rpm -iUh /root/download/*

# install php
yum install -y php

# install mysql system files
mysql_install_db --datadir=$mydir

# restart mysql
/etc/init.d/mysql restart

We can increase / decrease instances based on a particular image that will have all the data built-in. Here are the 4 commands those we can execute to create the auto-scaling group. Auto Scaling and Load Balancing tools can be installed from the following links...

http://aws.amazon.com/developertools/2535
http://aws.amazon.com/developertools/2536

Load balancer can be easily created / disabled from the web interface. But I guess for auto-scaling we will need to install the command line tool mentioned above.

1) Create a load balancer:
elb-create-lb mysql_LoadBal --listener "1b-port=3306,instance-port=3306,protocol=TCP" --availability-zones us-east-1b

2) Create auto scaling group launch configuration:
as-create-launch-config mysql_Config --image-id ami-60da3d09 --instance-type m1.small

3) Create auto scaling group:
as-create-auto-scaling-group mysql_AutoScale --launch-configuration mysql_Config --availability-zones us-east-1b --min-size 1 --max-size 5 load-balancers mysql_LoadBal

4) Create Trigger:
as-create-or-update-trigger mysql_Trigger1 --auto-scaling-group mysql_AutoScale --namespace "AWS/EC2" --measure CPUUtilization --statistic Average --dimensions "AutoScalingGroupName=mysql_AutoScale" --units "Percent" --period 60 --lower-threshold 30 --upper-threshold 70 --lower-breach-increment"=-1" --upper-breach-increment "1" --breach-duration 120

When average CPU utilization in the group exceeds 70% over a two-minute interval, increase instances up-to maximum of 5. When load falls below 30% reduce the number of instances but keep at-least 1 instance running.

To check how it is working:
as-describe-scaling-activities mysql_AutoScale
_____

i) In order to disable auto-scaling We will first have to remove the trigger:
as-delete-trigger mysql_Trigger1 --auto-scaling-group mysql_AutoScale

ii) Set minimum and maximum size of the group to 0 to force all its instances to terminate:
as-update-auto-scaling-group mysql_AutoScale --min-size 0 --max-size 0

iii) Remove auto-scaling group:
as-delete-auto-scaling-group mysql_AutoScale

iV) Remove load balancer:
elb-delete-lb mysql_LoadBal

Each EC2 instance can access run-time data about itself by making HTTP requests to the special address 169.254.169.254

wget -q http://169.254.169.254/latest/meta-data/instance-id

cat instance-id
i-d830dfb0

You can also supply your own metadata when you launch an EC2 instance. This data is called "User Data". For e.g.

wget -q http://169.254.169.254/latest/user-data
cat user-data
Role=master_db,Size=small,Name=Production,Input=Queue1

You can have your custom AMI treat the user metadata as the URL of the script to be run when the instance starts.

We can glue the EBS volumes together into a RAID.

mdadm --create /dev/md0 --level 0 --metadata=1.1 --raid-devices 2 /dev/sdh1 /dev/sdh2

Add the following line to /etc/mdadm.conf (creating if necessary)

DEVICES /dev/sdh1 /dev/sdh2

Run the following command to add additional configuration information to the file:

mdadm --detail --scan >> /etc/mdadm.conf

At this point the RAID volume (/dev/md0) has been created. It encapsulates the 2 volumes.

Create a file system:

mkfs /dev/md0

create a mount point and mount the volume on it:

mkdir /data
mount /dev/md0 /data

While using "system" function to execute linux commands within PHP, we can use a variable to store the return value. We will continue only if the return value is 0 (success) or exit the script.

system("mysqldump -uroot db1 --ignore-table=db1.tbl_1 --ignore-table=db1.tbl_2 > /root/db1.sql", $retval);

if ($retval != '0') {
  exit ("dump db1.sql command failed");
}

I have always ignored the following warning.

[Warning] Neither --relay-log nor --relay-log-index were used; so replication may break when this MySQL server acts as a slave and has his hostname changed!! Please use '--relay-log=slave-usw-relay-bin' to avoid this problem.

This made a big difference when I changed the AWS EBS and attached it to a new instance. Since the new instance had a new IP address, the relay-log file name was inconsistent. Therefore we must always specify log-bin and relay-log parameters in my.cnf

log-bin=usw-slave-bin
realy-log=usw-slave-relay-bin
_____

It is possible to recover slave if it has Failed to open the relay log. It is possible that MySQL saved it's relay logs in /var/run by default, which gets cleared out on boot.

To fix this, there are 2 ways.

you need to change the location MySQL uses for the logging by adding the following line to the [mysqld] section of /etc/my.cnf

relay-log = /var/lib/mysql/relay-bin

Then edit /var/lib/mysql/relay-log.info to point to the first new relay log (leaving the master information the same.

/var/lib/mysql/relay-bin.000001
1
mysql-bin.12345
123456789

Start the slave. Or if you already have the slave status at the time of stopping the slave. Note the following 2 parameters.

       Relay_Master_Log_File: mis-bin.000710
          Exec_Master_Log_Pos: 461352141

And then reset slave and start the replication by using the statement "CHANGE MASTER TO ..."