[S]cript [i]n [t]he [M]iddle.

SitM is a broker between zabbix-agent and AWS CloudWatch. Default list of CloudWatch metric is very limited. Well known problem that memory usage is a custom metric for every EC2 instance within AWS. Using SitM it’s easy to put any custom metric to CloudWatch.

Requirements.

• boto python lib
• zabbix-agent
• Python 3.x

Architecture.

Configuration

Main SitMs config file is /etc/sitm.conf

[common]
region = eu-west-1
instance-id = i-xxxxxxxxx
socket_timeout = 30
port = 10050
host = localhost

[mem-free]
namespace = MEM
unit = Percent
key = vm.memory.size[pavailable]

[metric-name]
namespace = group
unit = Count
key = something

Boto configuration. http://boto.cloudhackers.com/en/latest/boto_config_tut.html

How it works.

zabbix-agent is a place where we can take wide range of default metrics, plus any user defind metrics. Example:

# /etc/zabbix/zabbix_agentd.conf.d/redis.conf
---------------------------------------------------------
UserParameter=redis.db_size,/opt/redis/bin/redis-cli -r 1 -i 1 DBSIZE | awk '{print $1}'
UserParameter=redis.clients,/opt/redis/bin/redis-cli -r 1 -i 1 INFO | grep "connected_clients" | awk -F  ":" '{print $2}'
UserParameter=redis.req,/opt/redis/bin/redis-cli -r 1 -i 1 INFO | grep "instantaneous_ops_per_sec" | awk -F  ":" '{print $2}'

The only thing we have to do is to send request to zabbix-agent and put response to CloudWatch. Each non-[common] section of config represents single monitoring metric. Just create crontab file to start monitoring avaliable memory:

 */5 * * * *     zabbix    sitm.py mem-free
 */5 * * * *     zabbix    sitm.py metric-name
 ...