How to get filesystem via AWS EC2 dynamic inventory - amazon-ec2

Question

I found quite a specific issue when setting AWS CloudWatch alarms via ansible using the ec2 dynamic inventory.
I've successfully set up the aws-script-mon to monitor the usage of Disk and RAM usage on my machines.
Also I've managed to set RAM usage alarms with the ansible ec2_metric_alarm module.
The problem I'm facing at the moment is when setting the alarms for disk usage the Filesystem dimension parameter is required, but not returned in the ec2 dynamic inventory variables.
Some of my machines have filesystem set to /dev/xvda1 and others have something like: /dev/disk/by-uuid/abcd123-def4-....
My current "solution" is as follows:
- name: "Disk > 60% (filesystem by fixed uuid)"
ec2_metric_alarm:
state: present
name: "{{ ec2_tag_Name }}-Disk"
region: "{{ ec2_region }}"
dimensions:
InstanceId: '{{ ec2_id }}'
MountPath: "/"
Filesystem: '/dev/disk/by-uuid/abcd123-def4-...'
namespace: "System/Linux"
metric: DiskSpaceUtilization
statistic: Average
comparison: ">="
threshold: 60.0
unit: Percent
period: 300
evaluation_periods: 1
description: Triggered when Disk utilization is more than 60% for 5 minutes
alarm_actions: ['arn:aws:sns:us-west-2:1234567890:slack']
when: ec2_tag_Name in ['srv1', 'srv2']
- name: "Disk > 60% (filesystem /dev/xvda1)"
ec2_metric_alarm:
state: present
name: "{{ ec2_tag_Name }}-Disk"
region: "{{ ec2_region }}"
dimensions:
InstanceId: '{{ ec2_id }}'
MountPath: "/"
Filesystem: '/dev/xvda1'
namespace: "System/Linux"
metric: DiskSpaceUtilization
statistic: Average
comparison: ">="
threshold: 60.0
unit: Percent
period: 300
evaluation_periods: 1
description: Triggered when Disk utilization is more than 60% for 5 minutes
alarm_actions: ['arn:aws:sns:us-west-2:1234567890:slack']
when: ec2_tag_Name not in ['srv1', 'srv2']
The only difference between those two tasks is the Filesystem dimension and the when condition (in or not in).
Is there any way how to obtain the Filesystem value so I can use them as comfortably as let's say ec2_id? My biggest concern is that I have to watch filesystem values when creating new machines and handle lists of machines according to that values.

Answer 1

I couldn't found a nice solution to this probem, and ended up writing a bash script to generate a YAML file containing UUID variables.
Run the script on the remote machine using the script module:
- script: ../files/get_disk_uuid.sh > /disk_uuid.yml
Fetch the created file from the remote
Use the include_vars_module to import the variables from the file. YAML syntax requires hyphens be replaced with underscores in variable names. The disk label 'cloudimg-rootfs' becomes 'cloudimg_rootfs'. Unlabeled disks use the variable names: 'disk0, disk1, disk2...'
Scripts aren't ideal. One day I'd like to write a module that accomplishes this same task.
#! /bin/bash
# get_disk_uuid.sh
echo '---'
disk_num=0
for entry in `blkid -o export -s LABEL -s UUID`; do
if [[ $entry == LABEL* ]];
then
label=${entry: 6}
elif [[ $entry == UUID* ]];
then
uuid=${entry: 5}
fi
if [ $label ] && [ $uuid ]; then
printf $label: | tr - _
printf ' '$uuid'\n'
label=''
uuid=''
elif [ $uuid ]; then
printf disk$disk_num:
printf ' '$uuid'\n'
label=''
uuid=''
fi
done