Home assistant entities unknown

[depuits]

My home assistant was showing some old entities from drydock so i decided to delete the device and let the discovery add them again. After this my sensor are showing correctly and for all containers a new entity was created but all container entities stay on the status unknown.

Following is my environment on drydock:

  environment:
      DD_LOCAL_WATCHER: "false"

      DD_NOTIFICATION_MQTT_MOSQUITTO_URL: "mqtt://redacted:1883"
      DD_NOTIFICATION_MQTT_MOSQUITTO_USER: "redacted"
      DD_NOTIFICATION_MQTT_MOSQUITTO_PASSWORD: "redacted"
      DD_NOTIFICATION_MQTT_MOSQUITTO_HASS_ENABLED: "true"
      DD_NOTIFICATION_MQTT_MOSQUITTO_HASS_DISCOVERY: "true"

I am using agents to collect containers from other instances.

I guess I need to add something to trigger the notifications on the containers individually but I'm not sure what i should add. I was also trying to make some containers do an auto update but I haven't gotten that to work either. Not sure if those are related but the documentation share a lot between NOTIFICATION and ACTION.

[scttbnsn]

I found a topic mismatch that fits this: with agent topics enabled, HA discovery subscribes to an agent-specific topic while older Drydock builds publish container state on the unscoped topic. That is the default on 1.7; 1.6 only hits it with HASS_AGENTTOPICSEGMENT explicitly enabled. #1139 fixes it on the 1.8 development branch, and I am backporting it.

Can you share the controller/agent versions and one affected entity's discovery state_topic alongside the matching Drydock "Publish container result to" debug line? Redact credentials and private hostnames. You normally do not need extra notification labels for MQTT state. Auto-updates use separate action configuration, so the MQTT settings alone do not enable them.

[scttbnsn]

I have backport PRs up for both maintenance lines: #1203 for 1.6 and #1204 for 1.7. They reuse the shared topic construction from #1139, so discovery and state publishing agree while preserving each version’s existing defaults. Both passed the full local test gates, including 100% coverage. These are not released yet, and I still need your installed version to confirm which path you are hitting. The 1.6 hosted build also exposed a stale Alpine timezone-package pin, which I am fixing before the next candidate.

[depuits]

I am currently using version 1.7.0-rc.14

Following is the data from the topic homeassistant/update/dd_container_agent_swarm01_local_esphome/config

{
    "unique_id": "dd_container_agent_swarm01_local_esphome",
    "default_entity_id": "update.dd_container_agent_swarm01_local_esphome",
    "name": "esphome",
    "device":
    {
        "identifiers":
        [
            "drydock"
        ],
        "manufacturer": "drydock",
        "model": "drydock",
        "name": "drydock",
        "sw_version": "1.7.0-rc.14"
    },
    "icon": "mdi:docker",
    "entity_picture": "https://raw.githubusercontent.com/CodesWhat/drydock/main/docs/assets/whale-logo.png",
    "state_topic": "dd/container/agent/swarm01/local/esphome",
    "force_update": true,
    "value_template": "{{ value_json.image_tag_value }}",
    "latest_version_topic": "dd/container/agent/swarm01/local/esphome",
    "latest_version_template": "{% if value_json.update_kind_kind == \"digest\" %}{{ value_json.result_digest[:15] if value_json.result_digest else value_json.image_tag_value }}{% else %}{{ value_json.result_tag if value_json.result_tag else value_json.image_tag_value }}{% endif %}",
    "release_url": "https://github.com/esphome/esphome/releases/tag/2026.8.2",
    "json_attributes_topic": "dd/container/agent/swarm01/local/esphome"
}

The state_topic is not directly giving me results.

[scttbnsn]

Thanks, that confirms the version and discovery topic. The RC14 publisher uses dd/container/local/esphome for that container, while your discovery config points HA at dd/container/agent/swarm01/local/esphome. That matches the code defect fixed in #1204, which has now merged into dev/v1.7 after tests and review. It is not in a released candidate yet. No extra notification labels are needed for this fix. I will leave this open until we can verify the corrected state topic on a new build.

[scttbnsn]

The fix is shipped in v1.7.0-rc.16, with the release artifacts and both image architectures verified. Can you update to ghcr.io/codeswhat/drydock:1.7.0-rc.16 and check whether dd/container/agent/swarm01/local/esphome receives state and the HA entity leaves Unknown? No extra notification labels are needed. Old retained messages on the unscoped topic are not removed automatically.

The 1.6 backport is also published in v1.6.1-rc.14. Thanks for the repro. I am leaving this open until we confirm it on your setup.

[depuits]

Updating to v1.7.0-rc.16 fixes the unknown states.

[scttbnsn]

Thanks for confirming, @depuits, and for the repro that pinned this down. Closing this as fixed in v1.7.0-rc.16.