Home Assistant Says “Run Successful.” Why Is the Light Still Off?
Home Assistant Says “Run Successful.” Why Is the Light Still Off?
You’ve written an automation. You click “Run actions,” and the light turns on. But that evening, when you actually open the door, nothing happens.
The reverse can happen too: the trace shows that execution reached the turn-on-light step, with no obvious error, but the room is still dark.
These two hypothetical scenarios both look like an “unreliable automation,” but they call for checking different things. The scope of a manual test, the conditions during a real trigger, the target selected by the action, and the device’s actual response need to be examined separately, layer by layer.
This article is based on Home Assistant’s official documentation, marked as version 2026.11.1 at the time of review. It does not include hands-on testing with home devices. Menu names may change between versions; the key is to understand what each kind of test actually proves.
“Run actions” does not test the whole automation
An automation typically consists of triggers, conditions, and actions. For example: when a door opens and the sun has set, turn on the entryway light.
Selecting “Run actions” in the editor skips the triggers and the automation’s top-level conditions, and executes the action sequence directly. Conditions inside the action sequence still apply, such as checks within a choose branch.[1]
So, clicking it and seeing the light turn on proves only that this execution of the actions could turn on the light. It does not prove that opening the door will trigger the automation, or that its conditions will pass during a real run.
The official documentation also notes that this kind of run has no real trigger data or trigger ID. If an action branch depends on which trigger fired, a manual run may follow a different path from a real one.[1]
| Test method | What it mainly checks | What it does not also prove |
|---|---|---|
| Run one action on its own | Whether that action can execute under the current circumstances | Whether earlier variables and trigger data are correct |
| Run all actions | The action sequence and the conditions within it | Whether real triggers and top-level conditions work |
| Test one condition on its own | Whether that check passes at the current moment | Whether it will pass when the automation actually triggers |
| Let a real event trigger it | The complete path and device behavior for that run | That it will be reliable at all times and under all failure conditions |
Choosing the right test first can save you from changing a lot of configuration that was already working.
Find the trace for that run first, rather than looking only at the current state
Home Assistant traces record triggers, condition checks, actions, and related variables. The official troubleshooting documentation recommends starting with the trace.[1][2]
If the problem happened at 8 p.m., open the run from that time rather than looking at the green indicator beside a condition now. The fact that the sun has set now does not mean the condition passed at the time of the test.
First, confirm whether the automation started. If there is no corresponding run, check whether the automation is enabled and whether the input underwent the change it listens for. The official documentation explicitly states that an automation’s actions can still be run manually when the automation is disabled. A working button therefore does not prove that the automation is enabled.[2]
If it did start, follow the trace to see where the path ended. Stopping because a condition did not pass is a logic result; stopping at an action error calls for checking the target and parameters. Neither should be treated with a blanket “increase the delay” fix.
Also distinguish what the interface displays from the raw state. A door sensor’s interface may show “open,” while a state trigger actually compares against on. Putting the interface’s displayed wording directly into a state string may fail to match.[2]
No error sometimes just means no controllable light was selected
Suppose an action says to “turn on the lights in the study area,” but that area contains no light entities recognized by the action. The official troubleshooting documentation notes that a target with no matching entities may be skipped without an error.[2]
So even if the action finishes, check what it actually selected. A changed device area, a new label, or a switch mistaken for a light are all worth checking.
For diagnosis, you can temporarily run a small, focused test against one explicitly named light entity. If direct control of that light works but targeting it through an area or label does not, check the target set first. This is a way to narrow down the problem; it does not mean every final configuration should hard-code an entity.
If an entity is unavailable, turn instead to its power, connection, and integration status. Different integrations may handle this differently: some skip unavailable entities and log a warning, while others still send a command and then report an error.[2]
You still need to observe whether the light actually changes. Completion of an action call alone cannot prove that the physical light reached the expected state.
An overall run can finish even if something failed along the way
Home Assistant supports continuing to run actions after an error. The official documentation explains that when Continue on error is enabled, later actions can still run after an intermediate action fails, and the overall run may finish.[2]
Consider an example sequence: turn on a light, then send a notification. If execution continues after the light fails to turn on, receiving a “sequence finished” notification could easily make you think the light came on too.
In that case, inspect the error on the turn-on-light step rather than looking only at the final notification. A workflow that needs confirmation should also distinguish between “a request was sent” and “the target state was observed.”
Whether to continue should be designed around the dependencies between actions. Sending a fault notification after a light fails to turn on is useful; sending “lighting is on” after that failure would mislead the user.
Simulating a state is not just a test on paper
The official documentation provides ways to simulate state changes and events. These can supply trigger data to an automation and are useful for checking branches that depend on a trigger.[1]
But there is an important detail: changing an entity’s state in the state tool does not itself change the physical device. The resulting state change can, however, trigger real automations, which may then control real devices.[1]
Before simulating a door sensor opening, therefore, know what other rules listen to it. Do not inadvertently trigger an entire away-mode, alarm, or other linked sequence just to test the entryway light. You can start with an isolated test entity and test rule, then verify the real path once you know the scope.
After the simulation, the state shown in the interface may also temporarily differ from the device’s actual state until the device reports again. Mark the data as simulated in your test record so it is not later mistaken for a real sensor reading.
Finish troubleshooting with one complete record
A useful record can be simple: what you did and when, what the input entity reported, which automation triggered, which condition passed or failed, which entity was ultimately controlled, and whether the device responded.
Verify a single action first, then the conditions and branches, and finally run through the path with a real event. If you have identified an incorrect target, fix the target first. If you know the automation did not trigger, go back to the input; there is no need to replace the sensor and the network at the same time.
If the problem is that the light turns off while someone is still there, continue with auto-off timing and occupancy checks. The two troubleshooting processes can connect, but a misbehaving light does not mean they share the same cause.
“Run successful” is useful information. Troubleshooting is truly complete when you can explain why the light did or did not turn on in that particular run.
References
Review date: November 11, 2026.
- Home Assistant: Testing automations, step-by-step testing, running actions, trigger data, simulated inputs, and traces.
- Home Assistant: Troubleshooting automations, disabled status, raw states, empty targets, unavailable devices, and continuing after errors.
AI-generated concept illustration, not a test photograph or network topology diagram.
Discussion
Share a question or your experience. Comments appear after moderation. Your email is never displayed publicly. Comment privacy
Loading comments…