Hi @mpous,
So I tested out the configizer, it runs nicely, it connects to the device automatically and updates the config.json, it also makes it easier to read as before it was a messy looking json file whereas with the configizer it gets “pretty”, as in each key:value is on a new line, making it easier to follow.
I did change a couple lines in the run-one.sh:
This:
# don't run if the device has already been processed
grep -a -q '{} : DONE' config.log 2>/dev/null && exit 0
# these settings need to be handled separately because they need to be different
# per-device even when processing a large batch
newKey=$(grep "${uuid}" new_keys | sed 's/.*\s*->\s*//')
To:
# don't run if the device has already been processed
grep -a -q "${uuid} : DONE" config.log 2>/dev/null && exit 0
# Check if the new_keys file exists and process accordingly
if [[ -f new_keys ]]; then
newKey=$(grep "${uuid}" new_keys | sed 's/.*\s*->\s*//')
else
newKey=""
fi
As before, I would get the:
grep: new_keys: No such file or directory
If I wasn’t trying to change the API keys, hence there is no new_keys file, whereas with this update it won’t print that error. Not that it caused any issues but I just didn’t like how it showed that all the time.
I also made some changes in the config.sh code, mainly just for the udevRules option.
I changed the following function:
# Handling udevRules
udevrulesInsert() {
echo "Inserting udevRules values"
local TEMPWORK
TEMPWORK=$(tempwork)
for key in "${!UDEVRULES[@]}" ; do
jq ".os.udevRules.\"${key}\" = \"${UDEVRULES[$key]}\"" "$WORKCONFIGFILE" > "$TEMPWORK" || finish_up "Couldn't insert rule ${key}."
if [[ "$(jq -e ".os.udevRules.\"${key}\"" "$TEMPWORK")" == "" ]] ; then
finish_up "Failed to insert ${key} rule into config.json."
fi
mv "${TEMPWORK}" "${WORKCONFIGFILE}" || finish_up "Failed to update working copy of config.json"
done
}
To (Unsure why it won’t format as code in this post):
udevrulesInsert() {
echo "Updating udevRules values"
local TEMPWORK
TEMPWORK=$(tempwork)
# Iterate through existing keys and remove any that are not in the new configuration
existing_keys=$(jq -r 'try .os.udevRules | keys_unsorted[]' "$WORKCONFIGFILE")
for key in $existing_keys; do
if [[ -z "${UDEVRULES[$key]}" ]]; then
jq "del(.os.udevRules.\"$key\")" "$WORKCONFIGFILE" > "$TEMPWORK" || finish_up "Couldn't remove rule ${key}."
mv "${TEMPWORK}" "${WORKCONFIGFILE}" || finish_up "Failed to update working copy of config.json"
fi
done
# Add or update the new rules
for key in "${!UDEVRULES[@]}"; do
jq ".os.udevRules.\"${key}\" = \"${UDEVRULES[$key]}\"" "$WORKCONFIGFILE" > "$TEMPWORK" || finish_up "Couldn't insert rule ${key}."
if [[ "$(jq -e ".os.udevRules.\"${key}\"" "$TEMPWORK")" == "" ]]; then
finish_up "Failed to insert ${key} rule into config.json."
fi
mv "${TEMPWORK}" "${WORKCONFIGFILE}" || finish_up "Failed to update working copy of config.json"
done
}
In the main() I also changed:
This:
if (( ${#UDEVRULES[@]} > 0 )); then
DO_UDEVRULES="yes"
anytask="yes"
fi
To this:
DO_UDEVRULES="yes"
anytask="yes"
So that it executes the udevrulesInsert function whether or not there is a udevRule present.
This way, when you comment out or remove the UDEV rule you added, it will remove it from the config.json and leave behind:
“os”: {
“udevRules”: {}
}
This is just a quick change, in case somebody decides to use this, they can change/clean it up as they wish. 
If anyone will be using this with Git Bash on a windows machine, I just had to set GitBash to use the native SSH option:
> export PATH="/c/Windows/System32/OpenSSH:$PATH
Otherwise it wouldn’t ssh into the device. 
Other than that it worked great, for now I was just testing it on the devices that I can access here, just in case I broke anything.
If anyone is interested, I was adding this rule: Auto-Mount External Storage into the configuration of my device.
Also @jchristensen, thanks for suggesting the use of jq, as that helps to check for simple syntax issues if not wanting to use the configizer. I would suggest if you can SSH into the device, copy the config.json file, insert the changes you would like to make, test this with jq (for peace of mind) and then you can enter the relevant change using the configizer as this will restart the balena supervisor to push the updated config.json. As, I tested removing a comma from the udevrule that i was adding and it still pushed to the device without giving any error. Perhaps if itnerested, someone can integrate the jq check on the config file into the configizer itself.
To check the actual config.json file I simply ran the below code using powershell, as I was checking this on a windows machine:
jq . config.json > $null
echo $LASTEXITCODE
This will return an error for syntax if there is one, otherwise it will return 0.
Once I know there is no syntax issues I can SSH into the device, manually edit the config.json by copy pasting my updated config file. Otherwise, the configizer will make the changes for you.
