Memfault OTA for any Linux device

This guide will walk you through using the Memfault OTA API to update any Linux system.

If you are using a Yocto-based Linux distribution with SWUpdate, we recommend using the Memfault SWUpdate integration.

Introduction

Memfault provides a simple OTA API that can be used to update any Linux system. The API is designed to be simple to integrate and to work with any existing update mechanism.

The API is based on a simple HTTP request to the Memfault cloud. The request contains the current version of the device and the device type. The response contains the latest version of the device and an URL to download the update from.

Concepts

All OTA solutions will involve the same key components:

• A build system prepares the Update Payload.

  This can be a full-system image in tarball format, a raw ext4 image, or some more advanced format (e.g. delta updates). In most cases, this payload will be signed.

  This process should be automated and integrated into your CI/CD pipeline.

• A server distributes the Update Payload.

  The OTA server should support distributing different payloads to different devices (e.g. different Hardware Versions, different software versions for different groups, etc). To scale well, this server also needs to use a Content Delivery Network which will guarantee that updates can be downloaded as fast as possible from devices all around the world, and keep the total bandwidth cost low.

  This is one of Memfault's core offerings: we host the Update Payloads, give you precise control of who-gets-what, and use a CDN to distribute updates.

• An agent on the device will connect to the server and regularly poll for updates.

  When an update is available, the agent will notify an update installer who will verify the update and install it.

• The update installer writes the update to disk, notifies the bootloader that an update has been installed and reboots into the update.

• After rebooting, the device verifies it's operational and connected, and notifies the bootloader to stay on this update.

Some update systems, like SWUpdate and RAUC, will combine the agent and the update installer into a single binary. They will communicate to the server using the hawkbit protocol. This is fully supported by Memfault and documented in our OTA with SWUpdate guide. Using a separate agent to poll for updates provides more control over when they are installed.

For other systems (like a home-built update system for example; or the Mender open-source client), you can focus on building your installer and follow this guide to integrate your solution with Memfault OTA.

The following table illustrates how some common OTA solutions implement each common component . All of these cases use Memfault OTA as the OTA server and a custom script acts as the agent.

OTA Solution	Update Payload Name	Build System	Installer	Confirm
SWUpdate		bitbake swupdate-image	swupdate <image URL>	(not required)
RAUC	Bundle	RAUC Bundle class	rauc install <image file>	rauc status mark-good
Mender	Artifact	Create an Artifact	mender install <URL>	mender commit

tip

For a more in-depth look at all the parts that need to fit together to make OTA updates work, check out our post, Practical Introduction to OTA for Embedded Linux and the associated webinar.

Prerequisites

info

The following assumes you already have a Memfault Account. If you don't yet have an account, reach out to us!

To follow this guide, you will need:

• A Linux device with an internet connection
• A way to generate an Update Payload
• A command to install the Update Payload
• (Optional) A way to confirm that the update was successfully installed

Memfault OTA API

The Memfault OTA API is a simple HTTP API that can be used to check for updates.

It is available on https://device.memfault.com/api/v0/releases/latest/.

The API takes a few arguments:

• device_serial - This is used for determining the Cohort a Device is in.
• hardware_version - OTA payloads are specific to each hardware_version, so this is a required parameter.
• software_type - The software component on the device responsible for performing OTA updates. For most Linux projects this will be main.
• current_version (optional) - The current Software Version of the Software Type that performs OTA updates. This is used to determine if there is a newer Software Version available to be installed. If omitted, the latest version is returned for the device.

Example:

$ curl "https://device.memfault.com/api/v0/releases/latest?device_serial=DEMOSERIAL&hardware_version=evt1&software_type=main&current_version=0.0.1" \
  --header "Memfault-Project-Key: ${YOUR_PROJECT_KEY}"

If there is no new version available, the API will return a 204 No Content response.

If there is a new version, the API will return 200 with a JSON payload.

{
  "extra_info": null,
  "is_delta": false,
  "created_date": "2023-06-01T02:13:18.881435+00:00",
  "is_forced": null,
  "revision": "",
  "min_version": "",
  "reason": "latest",
  "version": "0.42.0",
  "artifacts": [
    {
      "extra_info": null,
      "sha256": "60e3c68ea2056aa2b587a3e4d621f373926a4e5471330658dc6f285a53476b03",
      "created_date": "2023-06-01T02:35:02.614996+00:00",
      "build_id": "",
      "file_size": 258048,
      "url": "https://ota-cdn.memfault.com/3327/6587/7718856587?token=kM8rUk8MfJJ2kJmBJyJj1jg7XMeFsqbB90qRSAYT3_U&expires=1689796800",
      "type": "firmware",
      "hardware_version": "qemuarm64",
      "md5": "551b386859663efe4935cd1fd955342e",
      "sha1": "75dcb80f5afde12ce0125c3fe0d7d7d733a8ed5b",
      "filename": "update-0.42.0.swu",
      "id": 14841
    }
  ],
  "must_pass_through": false,
  "display_name": "",
  "id": 11771,
  "notes": ""
}

tip

If all you need is the download URL, add /url to the API route and it will return only the download URL (or nothing for a 204 response).

Example OTA Agent

The OTA agent is responsible for polling the Memfault cloud for updates and installing them.

Polling for updates

The following example shell script polls the server and installs the update if one is available.

You will need to update update-installer with the command that installs the update payload.

/usr/bin/memfault-ota

#!/bin/sh
DATA_DIR=$(cat /etc/memfaultd.conf | jq -r '.data_dir')
PROJECT_KEY=$(cat /etc/memfaultd.conf | jq -r '.project_key')
SOFTWARE_TYPE=$(cat /etc/memfaultd.conf | jq -r '.software_type')
SOFTWARE_VERSION=$(cat /etc/memfaultd.conf | jq -r '.software_version')
eval "$(memfault-device-info)"

url=$(curl -f -s --get \
    --data-urlencode "hardware_version=${MEMFAULT_HARDWARE_VERSION}" \
    --data-urlencode "software_type=${SOFTWARE_TYPE}" \
    --data-urlencode "current_version=${SOFTWARE_VERSION}" \
    --data-urlencode "device_serial=${MEMFAULT_DEVICE_ID}" \
    --header "Memfault-Project-Key: ${PROJECT_KEY}" \
    "https://api.memfault.com/api/v0/releases/latest/url")
exit_code=$?

if [ "$exit_code" -eq 0 ] && [ -n "$url" ]; then
    # The following line will need to be updated for the specific update installer
    # you are using.
    update-installer $url
    if [ $? -eq 0 ]; then
      memfaultctl reboot --reason 3
    else
        logger -p user.err "OTA install failed"
    fi
else
    if [ "$exit_code" -ne 0 ]; then
        logger -p user.err "OTA status check request failed"
        logger -p user.err "curl returned exit code: ${exit_code}"
    fi
fi

A systemd service and timer can be used to run the script periodically:

memfault-ota.service

[Unit]
Description=Memfault OTA Service
Requires=network.target
After=network.target

[Service]
Type=oneshot
ExecStart=/bin/sh /usr/bin/memfault-ota

[Install]
WantedBy=multi-user.target

memfault-ota.timer

[Unit]
Description=Launch %i service every 6 hours

[Timer]
OnBootSec=3min
OnUnitInactiveSec=6h
Persistent=true

[Install]
WantedBy=timers.target

Confirming updates

If you also need to execute some command after the update installs and the device reboots, use something like the following script. It runs on boot and tries to contact the OTA server to confirm that the new firmware image has working connectivity.

You will need to update the exact commands used to confirm or rollback the update.

memfault-post-ota.service

[Unit]
Description=Memfault Post OTA Service
Requires=network.target
After=network.target

[Service]
Type=oneshot
ExecStart=/bin/sh /usr/bin/memfault-post-ota

[Install]
WantedBy=multi-user.target

/usr/bin/memfault-post-ota

#!/bin/sh
DATA_DIR=$(cat /etc/memfaultd.conf | jq -r '.data_dir')
PROJECT_KEY=$(cat /etc/memfaultd.conf | jq -r '.project_key')
SOFTWARE_TYPE=$(cat /etc/memfaultd.conf | jq -r '.software_type')
SOFTWARE_VERSION=$(cat /etc/memfaultd.conf | jq -r '.software_version')
eval "$(memfault-device-info)"

checkin () {
    status_code=$(curl -o /dev/null -I -L -s -w "%{http_code}" --get \
    --data-urlencode "hardware_version=${MEMFAULT_HARDWARE_VERSION}" \
    --data-urlencode "software_type=${SOFTWARE_TYPE}" \
    --data-urlencode "current_version=${SOFTWARE_VERSION}" \
    --data-urlencode "device_serial=${MEMFAULT_DEVICE_ID}" \
    --header "Memfault-Project-Key: ${PROJECT_KEY}" \
    "https://api.memfault.com/api/v0/releases/latest/url")

    if [[ "$status_code" == "200" || "$status_code" == "204" ]]; then
        # Update this for your specific OTA platform
        commit-update
        return 0
    fi
    return 1
}

until [ $RETRY_COUNT -eq 10 ] || checkin; do
    ((RETRY_COUNT++))
    sleep 10
done
[ $RETRY_COUNT -lt 10 ] && exit 0

# We have not been able to confirm connectivity with OTA servers - Rollback update
commit-rollback

# Firmware Update Error
memfaultctl reboot --reason 32777

Controlling when updates are installed

To control when updates are installed you can add checks in the script above that will immediately return from the update script if the device is not in a state where we would want to reboot.

You could also choose to install the update but not reboot until a later time.

The possibilities are endless and depend on your specific use-case. We are here to help guide you through this process.

Important Considerations

The Thundering Herd issue

Care should be taken to make sure that all the devices will not poll the server at the same time. This is especially important with large fleet of devices where thousands of devices may all connect at the same time and overwhelm the server.

note

The systemd example above runs 3 minutes after boot and then every 6 hours so it will not have this issue as long as the devices do not all boot at the exact same time.

Common causes of this issue are:

• All devices are configured to poll the server at a fixed time (for example with a CRON job that runs every hour).

  Solution:

  • Avoid using CRON to run the agent on device
  • Randomize the polling time on each device
  • Add a long-sleep at the beginning of the CRON job

• All devices are configured to poll the server on boot and they are all restarted at the same time.

  Solution:

  • Wait a random amount of time before polling the server on boot
  • Avoid restarting all your devices at the exact same time

Polling Rate

We generally account for a maximum of 6 device checkins per day per device.

For devices in developer mode we will allow an unlimited number of checkins.

For other devices, they may run into rate-limiting issues if they check in more frequently and will receive a status code 429 Too Many Requests. Please contact us if your use-case requires a higher limit.

Security

We have kept security considerations out of the scope of this document because security is generally implemented in the update installer.

caution

It's extremely important to sign the update payload, and to verify the signature before installing it. If the update is not signed, an attacker with physical access to the device, or just the ability to alter the network around the device, could trick the device into downloading from a malicious server and install an update which would give the attacker complete control on the device.

All commercial and open-source solutions support signing updates and will provide specific documentation on how to do it.

For reference:

• SWUpdate - Signed Images
• RAUC - Security documentation
• Mender - Sign & Verify Guide