v2.1.0

Major updates

Extensions

This release combines with nerves_hub_link v2.6.0 to bring the complete Extensions experience to NervesHub. An Extension is a non-critical but useful mechanism that needs to piggy-back on the device's connection to NervesHub. Rather than opening more connections or risking that the critical firmware delivery socket fails we have wrapped both ends of the socket in safeties.

Extensions are only enabled after the essential connection is confirmed and firmware delivery is possible. Each extension is also opt-in at the product level under product settings and possible to opt out at the device level under device settings. These changes can importantly be made at run-time in the UI. We want to ensure that if anything is wrong with an extension or it somehow disrupts critical operation it is trivial to disable immediately. This is to ensure that firmware updates and trouble-shooting come first.

The upshot of this is that we can build some very nice features without touching the core of what NervesHub must do. More on those below.

Extension: Health

Health reporting lets your device send metrics, metadata and alarms to NervesHub giving you indicators about your device fleet's health. Health information is shown on the device details page when the extension is enabled and as information gets reported.

Default metrics, if supported by the hardware:

• CPU temperature
• CPU utilization
• Load averages (1, 5, 15 min respectively)
• Memory usage
• Disk usage

Adding custom metrics to the default report is a minimal addition in config.exs on nerves_hub_link:

config :nerves_hub_link,
  health: [
    # metrics are added with a key and MFA
    # the function should return a number (int or float is fine)
    metrics: %{
      "battery_percent_remaining" => {MyBattery, :percent_remains, []}
    }

There are many more features planned for metrics. Filtering, monitoring levels and various types of acting on metrics being out of bounds to provide a good idea about device health.

Alarms are a very nice way to find out if your devices is experiencing issues and uses the Erlang :alarm_handler functionality. The default alarm handler is not intended for real use and when it is used we won't report alarms because they don't behave very well. The alarm handler we know well and can recommend is alarmist.

The device list can be filtered by devices with any alarms, specific alarms and your navigation bar within a product will indicate devices with alarms helping you quickly spot and find the list of devices with active alarms.

Extension: Geo

Geo allows your device to report location information. By default it uses a Geo-IP mechanism through an API call to a web service. This gives a rough location. A device with location information can be shown on the Dashboard map which is enabled by setting DASHBOARD_ENABLED=true. If you configure a Mapbox API key via the environment variable MAPBOX_ACCESS_TOKEN it will fetch a local map for the device details page as well.

There are certainly more things we want to do in terms of filtering and useful maps with this geo-location information and this is the starting point.

Your device may well have a better way to get location data, whether LTE, GPS or manually entered coordinates you can modify the device's config.exs to include:

config :nerves_hub_link,
  geo: [
    # Implement your  `NervesHubLink.Extensions.Geo.Resolver`, just one function
    resolver: MyGPS.Resolver
  ]

Added

• Add the oban integration to Sentry (#1543)
• Add container version tagging (#1549)
• Add device count and estimated device count for Deployment views (#1517)
• Use Chart.js for metrics (#1523)
• Add task for generating randomized device metrics (#1564)
• Add support for a release id for Sentry (#1568)
• Add option for showing updated devices on the map (#1592)
• Clear inflight updates for unhealthy devices (#1620)
• Show helpful message when uploading duplicate firmware (#1627)
• Add support for filtering devices on alarms (#1628)
• Have the calculator use Oban (#1639)
• Save device connection data over time (#1572)
• Add support for OpenTelemetry tracing (#1612)
• Show current alarms on device page (#1648)
• World Map clustering (#1619)

Changed

• Device channel cleanups (#1546)
• Increase the Repo queue_target (#1545)
• Fix dialyzer warning about use of map where each would do (#1550)
• Use the Endpoint from the socket (#1558)
• Sentry integration tweaks (#1560)
• Remove the NodeReporter metrics logger (#1563)
• Only render connection tooltip when timestamp is present (#1565)
• Logger improvements (foundations) (#1556)
• Allow errors and warnings in test env (#1561)
• Separate logging from metrics (#1571)
• Allow the socket drainer to be configured at runtime (#1577)
• Use the endpoint from the socket (#1579)
• Reintroduce missing custom metrics and metadata to device page (#1578)
• Add y-scale start number as chart parameter (#1580)
• Metrics charts improvements (#1581)
• Only broadcast on terminate/2 if the channel is joined (#1586)
• Display selected option when moving device(s) to other product (#1591)
• Add firmware delta generation back in (#1582)
• Order platform dropdown + add unknown platform as selection (#1599)
• Group by deployment id when counting devices for deployments (#1600)
• Disable console button when device console is not available (#1605)
• Batch insert device metrics (#1613)
• Optimize the queries needed for Accounts.list_org_keys (#1614)
• Increase the Orchestrator timer, and add a jitter (#1615)
• Remove an extra Deployment preload during after_join (#1618)
• Allow publishing docker images from PR (#1626)
• Use PRs HEAD sha for tagging images (#1637)
• Use PR number when looking for publish commit message (#1638)

Fixed

• Some minor dialyzer fixes (#1553)
• Update mix.exs Elixir version to match .tool-versions (#1621)
• Use the correct endpoint for pubsub broadcasts (#1634)
• Fix an unreachable with/else code path (#1642)
• Use the full module reference for some @specs (#1645)
• Change health section to not show boxes for non-reported default values (#1651)

Dependencies

• castore: 1.0.8 -> 1.0.10
• ex_aws_s3: 2.5.3 -> 2.5.5
• dialyxir: 1.4.3 -> 1.4.5
• bcrypt_elixir: 3.1.0 -> 3.2.0
• comeonin: 5.4.0 -> 5.5.0
• swoosh: 1.17.1 -> 1.17.3
• slipstream: 1.1.1 -> 1.1.2
• telemetry_metrics_statsd: 0.7.0 -> 0.7.1
• ecto: 3.12.3 -> 3.12.4
• ecto_sql: 3.12.0 -> 3.12.1
• telemetry_metrics: 0.6.2 -> 1.0.0
• ecto_psql_extras: 0.8.1 -> 0.8.2
• ex_aws: 2.5.5 -> 2.5.7
• x509: 0.8.9 -> 0.8.10
• postgrex: 0.19.1 -> 0.19.3
• floki: 0.36.2 -> 0.36.3
• crontab: 1.1.13 -> 1.1.14
• sentry: 10.7.1 -> 10.8.0
• phoenix_ecto: 4.6.2 -> 4.6.3
• gettext: 0.24.0 -> 0.26.2
• phoenix_live_dashboard: 0.8.4 -> 0.8.5
• bandit: 1.5.7 -> 1.6.0
• opentelemetry_bandit: 0.2.0-rc.1 -> 0.2.0
• opentelemetry_phoenix: 2.0.0-rc.1 -> 2.0.0
• open_telemetry_decorator: 1.5.7 -> 1.5.8

From assets/:

• webpack: 5.45.1 -> 5.95.0
• braces: 3.0.2 -> 3.0.3

Full Changelog - v2.0.0...v2.1.0

v2.0.0

Trying to create a Changelog from the previous tagged version v1.3.0 from
January 2023 is frankly a bit much.

The fundamentals:

• Performant at scale: It is used in production with hundreds of thousands of connected devices. It works well.
• Remove AWS: It is now easy to run and set up on any hosting platform

There is also a lot more. The initial announcement of 2.0
has the basic idea. The only idea that isn't currently actively pursued or
already delivered is MQTT and that is bound to pop up again in the future.

A lot of this work should be credited to Eric Oestrich and he gave an update
on it at the Nerves Meetup.

Jon Carstens covered some of the work during
his NervesConf talk as a well.

After the 2.x changes were available Josh Kalderimis joined in and started
looking at ease of use which led to a lot of various improvements but the
biggest one is the alternative device authentication method: Shared Secret

Since then we have contributions from 19 people where Eric Oestrich dominates
the stats. Followed by Josh Kalderimis. Then Jon Carstens.

There are a lot of things planned. SmartRent, NervesCloud and other users of
the project are in active collaboration on improving the featureset, the
performance and the user experience. We want this to be not just a great tool
for developers that choose Nerves but a surprising delight and joy to use.

Migration

You should be able to migrate using the Ecto migrations. Nothing should be
break if you migrate a 1.x install to the 2.x releases. We still recommend
that you set aside time for dealing with anything unusual that might come up.

You may also need some time to deprovision the AWS resources you no longer
need.

v1.3.0

Potentially Breaking Change

Versions were updated to Elixir 1.14.2 and OTP 25.2. Much testing has been done to
attempt to catch any potential SSL/TLS issues before release and the update should
be fairly safe, but it is worth keeping an eye on.

Added

• Add NervesHubWebCore.Workers.TruncateAuditLogs for periodic cleaning of
  audit logs table (thanks @LostKobrakai)
• Firmware UUIDs hyperlink to the firmware#show page (thanks @TheCraftedGem)
• Add pretty 404 Not Found page (thanks @TheCraftedGem)
• Audit Log on device disconnect (thanks @TheCraftedGem)

Fixed

• TLS 1.2 is forced for support with past OTP versions and to prevent
  devices using cryptochips from being able to connect
  • Note: :sha and :sha224 signature algorithms were dropped as there is a potential
    bug negotiating them on a device with cryptochips if the server presents them as options.
    Since they are not typically used, it was decided to remove the support to fix the bug
    until more investigation can be done when reviewing OpenSSL 3.0
• [#871] Fix JITP>product relation to allow multiple profiles (thanks @jeanparpaillon)
• DeviceLive.Index now sorts and paginates via the database instead of loading
  all devices into memory. Fixes an issue where a production with thousands of devices
  may fail to load the index page (thanks @oestrich)
  • Moving to the DB level broke the ability to sort/query by Connection Status.
    This will be adjusted and fixed in a later release
• Fix incorrect query in Fix Accounts.get_user_by_email_or_username/1 (Thanks @zolakeith!)
• Paginate audit logs instead of loading the complete feed (thanks @LostKobrakai)
• Sort devices Last Communication correctly when nil

Updated

• Device imports now support certificates Base64 encoded as DER format
• Sort firmware dropdown options by version number (thanks @TheCraftedGem)

v1.2.0

v1.2.0 - 2022-10-18

Potentially Breaking Change

• Moved flag for enabling/disabling Delta updates to the Deployment instead of
  the Product. This allows for more granular control over which devices get
  Delta updates. This will require migrations to be run before use.

v1.1.0

Fixed

• Default to TLS 1.2 for all connections. This fixes issues if TLS 1.3 is
  used or attempted. See NervesHubWeb: Potential SSL Issues

v1.0.4

Fixed

• Fix return value passed to consume_upload_entry/3 (thanks @tonnenpinguin)

v1.0.3

Fixed

• [#844] Removed JoshJS to fix bug preventing some pages from loading (:heart: @pojiro)
• [#847] settings/<org>/certificates would crash if certificate had a JITP profile (:heart: @pojiro)

Updated

• JavaScript library updates
  • moment 2.29.2 -> 2.29.4
  • terser 5.12.1 -> 5.14.2

v1.0.2

Fixed

• [#838] Unregistered, expired signer CAs in the device connection request
  would fail before the device certificate pin could be checked

v1.0.1

Fixed

• Fixed on/off toggling on deployment#show page (thanks @tonnenpinguin!)

v1.0.0

The official start of NervesHub versioning. See the NervesHub documentation
for more information and Custom Deployment
guide for setting up your own instance.