Automate boring configuration changes

Sometimes we do repetitive actions that seem not worth to automate. The problem is that sometimes these affect morale and become time-consuming: for instance for lack of attention the configuration change introduces regressions, which cause an extra deployments to fix errors.

Well I had some time to wonder if I could automate adding entries to our one of our configuration file in a consistent way. And as a side effect I improved my Bash scripting skills.

The configuration file I am speaking about whitelists applications. Our configuration file contains entries which represent applications. Each entry has an unique identifier. Anytime an application requests a service, it has to submit this identifier: the request is fulfilled only if the identifier is in the configuration file.

When we insert a new entry manually, we have to satisfy two constraints:

we have to check that the new identifier is not already present
we have to add the entry to multiple configuration files (each representing a different deployment environment)

So our Bash script will do the following sequentially:

ask required input
generate an uuid
check if the generated uuid already exists and generate a new one until it does not
make the entry
append entry to the files

So here is the script:

## Usage: the script will request the necessary inputs to create a new application entry and will alter the configuration files in the repository.

files="some-dir/resources-local/configurationFile.json some-dir/resources-DEV/configurationFile.json some-dir/resources-TST/configurationFile.json some-dir/resources-ACC/configurationFile.json some-dir/resources-PRD/configurationFile.json"
files_array=($files)
prd_file=${files_array[${#files_array[@]}-1]}
# ask for user input
read -p 'Application name: ' appname
read -p 'Type: ' type
read -p 'Contact name: ' contact_name
read -p 'Maintainers email address: ' email
read -p 'Aim of the application: ' description
echo Current vars: $appname $type $email $description

# generate a new lowercase uuid (it must not be already in the file)
uuid=$(uuidgen)
lowercase_uuid=$(echo $uuid | tr '[:upper:]' '[:lower:]')
echo New uuid: $lowercase_uuid
# retry if uuid already exists
while true; do
    if grep "$lowercase_uuid" $prd_file > /dev/null
    then
        uuid=$(uuidgen)
        lowercase_uuid=$(echo $uuid | tr '[:upper:]' '[:lower:]')
        echo 'new uuid found in file, created a new one: ' $lowercase_uuid
    else
        echo uuid not found
        break
    fi
done
# end of retry

# make a new configuration entry
entry='
{\n
  "contactName": "'$contact_name'",\n
  "contactEmail": "'$email'",\n
  "description": "'$description'",\n
  "id": "'$lowercase_uuid'",\n
  "name": "'$appname'",\n
  "type": [
    "'$type"
  ]
}'

echo $entry

for file in "${files_array[@]}"
do
    # remove blank lines at the end of the file
    sed -i '' -e :a -e '/^\n*$/{$d;N;};/\n$/ba' $file
    # remove last line (because it contains "]")
    sed -i '' -e '$ d' $file
    # remove blank lines at the end of the file
    sed -i '' -e :a -e '/^\n*$/{$d;N;};/\n$/ba' $file
    # remove last line (because it contains "}")
    sed -i '' -e '$ d' $file

    # append entry
    echo '},\n' $entry '\n]' >> $file
done

This kind of script is not perfect, because it works on assumptions like:

the Production configuration file contains all the entries identifiers
the user inputs valid information
indentation of the json files is irrelevant

However, this script made a configuration change a thoughtless process on my side. The only automation left is to ask people needing an identifier to open a merge request to the repository by using this script.

Maybe this can inspire you to hack (better than this) some small and frequent&boring tasks you have to do.

Happy hacking!

Where parallels cross

Automate boring configuration changes

Comments