[Support]: Some events 'disappear' #19135
Replies: 4 comments 4 replies
-
I'd be very surprised if this was some kind of Frigate bug, as something as severe as "missing" events would be much more highly reported considering Frigate's userbase. There are a few important points to consider in your use case:
There's more I could add, but those are the first few things that are the most important to check. |
Beta Was this translation helpful? Give feedback.
-
Thanks for the quick response.
One thing I have checked is the false positive flag in the MQTT messages, and it is always false, except for the first message with the type of new.
I have also noticed that the has_snapshot and has_clip fields are also always false (in the mQTT events message) - yet a snapshot is returned from the API (when queried which the event is active).
From your second bullet - does this mean the has_snapshot field is only really valid when the event has finished and a snapshot has been saved? I have not checked that - I can query the API for event details for one of my missing events and see if it exists - I have only tried to grab the snapshots upto now.
Might be a couple of days before I can update as I am away from home.
…________________________________
From: Josh Hawkins ***@***.***>
Sent: Monday, July 14, 2025 2:10 PM
To: blakeblackshear/frigate ***@***.***>
Cc: Chris Henderson ***@***.***>; Author ***@***.***>
Subject: Re: [blakeblackshear/frigate] [Bug]: Some events 'disappear' (Discussion #19135)
I'd be very surprised if this was some kind of Frigate bug, as something as severe as "missing" events would be much more highly reported considering Frigate's userbase.
There are a few important points to consider in your use case:
* The frigate/events MQTT topic is a raw topic. False positives and tracked objects that will not be retained are reported with this topic. It's the user's responsibility to filter these out (by checking false_positive, for example).
* The snapshot endpoint will always return a valid snapshot for in-progress objects, but as soon as the object's lifecycle is ended, there may not be a snapshot immediately available. has_snapshot is set to true when a frame is saved. The snapshot response is generated on the fly until the object stops being detected at which point they are written to a file.
There's more I could add, but those are the first few things that are the most important to check.
—
Reply to this email directly, view it on GitHub<#19135 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ASRHCNSBBN5DQI3FGXLTLPD3IMGSVAVCNFSM6AAAAACBNTKWBKVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTGNZUGY4TONA>.
You are receiving this because you authored the thread.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
I have just done a quick check on the API - to get the event details - not try and get a snapshot. http://172.21.64.125:5000/api/events/1752438631.822133-joq55d This is the same event as in the logs section about with the MQTT messages - and at no point was it marked as a false positive. |
Beta Was this translation helpful? Give feedback.
-
ok. makes sense. i can check the position changes easily enough. I guess a valid question - is the current behaviour correct? There definitely was a cat in the example above. The percentage of these is not insignificant. Is it worth me checking more examples to see if the detected object was there or not - or is the feeling that changing the current behaviour would cause a significant number of false positives? |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Checklist
Describe the problem you are having
Not going to be an easy one to troubleshoot - but I thought I would raise an issue, see if it known behaviour.
I have written a node-red flow that listens to the frigate events. It subscribes to the MQTT event messages, and then sends some notifications based on some rules. It listens to each MQTT message for an event, through the new, update, end events. At each event it logs to a database, and also grabs a snapshot (and thumbnail) from the HTTP API. For the new event data is taken from the before section of the message, for each update and end data is taken from the after section. All of this is logged to a database, and snapshots are saved externally to frigate.
Sometimes (maybe 5% of events), once the end update is reached the snapshot is not available. I know that the object that was detected is actually there - as I can see it in the early JPGs. The flow will try and grab a snapshot within a few milliseconds of the MQTT message. Interestingly if I query the snapshot URL through a browser some time later then the entire event id seems to have gone.
Steps to reproduce
I can't reproduce this at will.
Version
0.16.0 latest
In which browser(s) are you experiencing the issue with?
not a browser issue i think
Frigate config file
docker-compose file or Docker CLI command
Relevant Frigate log output
Relevant go2rtc log output
Operating system
Other Linux
Install method
Docker Compose
Network connection
Wired
Camera make and model
axis various models
Screenshots of the Frigate UI's System metrics pages
not sure this will be needed. server is well spec'ed and does not seem to have any performance issues. running on a tesla gpu.
Any other information that may be helpful
can provide other information as required.
Beta Was this translation helpful? Give feedback.
All reactions