Closed apedance closed 2 months ago
Duplicate of #117 I think
Indeed. But not for the media part. Not having that issue. Just the reaching limit of channel. Any idea how to solve?
edit: in the making. refactor.
Right but the (ID: 3) is a media message so it should be the same thing
As far as I can tell it is because the media part of the channel is not pulling the messages off of the stream fast enough. The whole part of the media code is quite complex so I am working on a refractor to hopefully remove this complexity and make it easier to pull messages off of the queue
I have 6 cams on Neolink. I've tried Windows environments, Linux. The consistent theme I've come to is that they work much better if you only have 1 cam per instance. I get far less drop-outs on Frigate and the Frigate logs are clear. I've now separated out a Debian LXC on proxmox for each cam. Once you have the first one established, just copy it, edit TOML and away you go. They've only been up a day on their LXC, but they seem to behave very differently in terms of their resource consumption. Some will gradually fill the memory on the LXC throughout the day, some are quicker. Is there any recommended specs? I'm running each container with 512GB mem and 2VCPU. The VCPUs sit at next to nothing (under 3%) generally
This should now be fixed on latest, can anyone confirm?
Hi. Same error for me, with the build you sent me to fix rtsp error 404. After few hours of correct functioning i have this error.
it seems to be very similar to the previous timed out error, with the difference that some cams present the rtsp 503 error
then trying to close and reopen neolink, a series of gstreamer errors appear.
So this is quite odd. This warning message occurs when 100 messages are in the queue. Meaning we got 100 messages from the camera but the app has not processed them.
The odd thing is that there is no real processing going on. The app pulls a message converts it to bcmedia then instantly puts it in a broadcast stream with no waiting around.
The only part I can think of that might be causing this is if the conversion from bc->bcmedia is somehow stalling the process. So in that regards I made this build that uses a different (and hopefully less error prone) way of filtering and converting bc packets into bcmedia packets. Could you test it?
So this is quite odd. This warning message occurs when 100 messages are in the queue. Meaning we got 100 messages from the camera but the app has not processed them.
The odd thing is that there is no real processing going on. The app pulls a message converts it to bcmedia then instantly puts it in a broadcast stream with no waiting around.
The only part I can think of that might be causing this is if the conversion from bc->bcmedia is somehow stalling the process. So in that regards I made this build that uses a different (and hopefully less error prone) way of filtering and converting bc packets into bcmedia packets. Could you test it?
Sure, as soon as I have time today I will do it and I'll update you!
So this is quite odd. This warning message occurs when 100 messages are in the queue. Meaning we got 100 messages from the camera but the app has not processed them.
The odd thing is that there is no real processing going on. The app pulls a message converts it to bcmedia then instantly puts it in a broadcast stream with no waiting around.
The only part I can think of that might be causing this is if the conversion from bc->bcmedia is somehow stalling the process. So in that regards I made this build that uses a different (and hopefully less error prone) way of filtering and converting bc packets into bcmedia packets. Could you test it?
I can't download it, it says it doesn't exist
Seems I missed the 0
at the end of the link. Here is the right one
Seems I missed the
0
at the end of the link. Here is the right one
Tested, all cam seems to work but just started it gives errors in gstream
Ah that error, it was part of my experiment with buffering in gstreamer, it is already fixed in latest
I think you can use this build to confirm its fixed
I think you can use this build to confirm its fixed
This build is being tested and hasn't given any errors so far, but it seems (particularly in the mainstream) to give this saturation type effect. I don't know if it could depend on the FPS, which as I was saying, often drops dramatically to 10 or less.
This isn't saturation it's an incomplete video frame. These usually happen across reconnect boundaries or at the start of a stream
What I can do is make neolink wait for a key frame before sending anything onwards between resets but this will mean that some frames will be skipped while waiting for a key frame
So build with the filtered for first keyframe is here at this point though perhaps we should test if these lots of reconnects are happening because of network congestion. Have you tried things like reducing the bitrate and fps on the cameras?
So build with the filtered for first keyframe is here at this point though perhaps we should test if these lots of reconnects are happening because of network congestion. Have you tried things like reducing the bitrate and fps on the cameras?
Good morning. the latest build I was testing this morning seems to have failed. I didn't find any errors, the neolink process simply closed. a bit strange. However the bitrate is 4096 on the main stream with 20 fps
I am not observing this close without errors issue that you have. Also perhaps we should do this on another thread since it seems to not be related to this issue
I am not observing this close without errors issue that you have. Also perhaps we should do this on another thread since it seems to not be related to this issue
Ok, I will open another thread
Hi! same problem with the latest build you posted too @QuantumEntangledAndy
That's a shame how long does it take to get to this state? Do you still get to it even with 1 camera? I ask because if I want to replicate and trace this issue I might need to make a wireshark dump but I only have 2 cameras and wireshark won't play nicely over hours of connectiity
Does this error correct itself? I have a watch dog that looks for 3s of no video and then restarts it. This should look like 3s of no video for the watch dog which should mean that the stream will restart and you will get 100 message space again
If you can turn on debug log you can see when the watchdog kicks the stream. Set the env varRUST_LOG
to debug
That's a shame how long does it take to get to this state? Do you still get to it even with 1 camera? I ask because if I want to replicate and trace this issue I might need to make a wireshark dump but I only have 2 cameras and wireshark won't play nicely over hours of connectiity
I didn't count it exactly, but the time is certainly more than 6 hours of correct functioning. There seems to have been a slightly different error, still relating to this typology.
However, no, it doesn't seem to resolve itself. I always have to restart neolink.
could I try creating 2 separate instances with different ports, one with 8 cams and one with just one cam, to see what happens? @QuantumEntangledAndy
It look like you ran out of RAM could you check that? Maybe show a graph of ram over time
It look like you ran out of RAM could you check that? Maybe show a graph of ram over time
I haven't had a chance to check these days. I think I'll be able to let you know tomorrow. Has 0.6.1 fixed any of my problems from what you know? THANKS! Mc @QuantumEntangledAndy
0.6.1 is mostly an mqtt fix but it does include the other fixes we did with regards to BI start ups. You can see a summary of what's included in the release page
It look like you ran out of RAM could you check that? Maybe show a graph of ram over time
testing with 0.6.2 Everything seems to work fine for the first 9 hours. However, I notice that the neolink task manager seems to consume more and more RAM, to restore it just restart the process. It is currently consuming around 5GB and increasing more and more
Yes I got rid of the large mem growth in 0.6.2 but there is a little bit left. I have some more fixes in the work that takes it down to zero at least for my macOS and Linux (can't test the windows) hopefully that will fix it.
Yes I got rid of the large mem growth in 0.6.2 but there is a little bit left. I have some more fixes in the work that takes it down to zero at least for my macOS and Linux (can't test the windows) hopefully that will fix it.
Hi! @QuantumEntangledAndy
unfortunately the previous problem seems to have recurred again with 0.6.2 master, I only notice that the error starts appearing after many more hours than in previous versions and that a greater quantity of errors appears. The symptom is a general lowering of the cam fps. Unlike other builds which completely stopped working. I don't know if this could depend on the RAM which is gradually being occupied more and more.
I think this might be partially on the hardware side too. Neolink drops the camera when this starts happening, yes, but I even see a lot frame dropping and general instability on the stream through reolink too, only on the one device so far. E1 if it helps.
Hi, are you guys still facing this issue?
I'm using a Reolink Lumus and seeing simliliar issues here:
[2024-01-09T23:35:52Z WARN fcm_push_listener::listener] Connection failed. Retrying. [2024-01-09T23:39:32Z WARN neolink_core::bc_protocol::connection::bcconn] Reaching limit of channel [2024-01-09T23:39:32Z WARN neolink_core::bc_protocol::connection::bcconn] Remaining: 100 of 100 message space for 5 (ID: 3) [2024-01-09T23:39:39Z WARN neolink_core::bc_protocol::connection::bcconn] Reaching limit of channel [2024-01-09T23:39:39Z WARN neolink_core::bc_protocol::connection::bcconn] Remaining: 69 of 100 message space for 5 (ID: 3) Killed
Haven't had any reports on this error in quite some time, so I will be closing it. Let me know if this is still happening on latest
Describe the bug neolink camera stops working after x amount of time with error:
[2023-08-28T22:44:50Z WARN neolink_core::bc_protocol::connection::bcconn] Reaching limit of channel [2023-08-28T22:44:50Z WARN neolink_core::bc_protocol::connection::bcconn] Remaining: 0 of 100 message space for 4 (ID: 3)
hard reset of neolink is needed to get the camera working again.