All was working fine for me streaming my Hikvision cams to my Google Hubs until 110 or possibly 109. Nothing has changed with my camera setup, so I’m trying to narrow down the cause. I see that mDNS needs to be allowed for cast devices now, which I am. But other than that, the only hint I’ve seen is:
2020-06-07 14:54:47 INFO (MainThread) [homeassistant.components.stream] Started stream: rtsp://hass:[email protected]:10554/Streaming/Channels/801
2020-06-07 14:54:47 ERROR (stream_worker) [libav.h264] reference picture missing during reorder
2020-06-07 14:54:47 ERROR (stream_worker) [libav.h264] Missing reference picture, default is 2
2020-06-07 14:54:47 ERROR (stream_worker) [libav.h264] number of reference frames (1+2) exceeds max (2; probably corrupt input), discarding one
But not sure if that’s new or not. All I get now is the black screen with “Smart Home Camera” on it that times out. Going to try rolling back here, but was curious if anyone is having success in 110.x ?
I’m streaming UnifiProtect RTSP streams without issue on 0.110.x but I did upgrade FFmpeg to v4 manually before updating HA having reviewed the release notes discussion thread.
There are details in the integrations page that may apply for you also.
Note that I don’t have VLANs etc configured, just a flat network.
I actually figured out my problem late last night. It involved the spit of base_url (which I was using) into external_url and internal_url in 110. Once internal_url was setup properly for my proxy, I was back in business.
I run Home Assistant (and Hass.io) via Docker on a home server. I have it configured to run on 127.0.0.1:9021 and then expose it through an nginx TLS proxy. I also block public DNS for my Google Hubs so they are forced to use my internal DNS server, which then resolves my HA domain name to a local IP rather than letting it loopback through the router.
Previously I was using base_url to specify that exposed secure url, but had migrated that to only external_url in 110 when base_url was split into two. So I think HA’s best guess of my internal_url was probably unreachable on port 9021. Setting that resolved things.
Yeah, I am trying to find out, why I have so much buffer if I use the hub, I only have nabucasa…
I think my stream uploads to whatever Google location, to download again afterwards…
I don’t have an internal base url, since that obviously is not going to work for Google, it will never reach it
Well, the only way I know of keep the streams direct on the LAN is to force the Google Hubs to use private DNS service that resolves the stream addresses to be local. I know in my case nothing touches the router for the streams, and they perform pretty well at 1080p/30. But I’m not sure how that could really be done w/ Nabucasa off hand. Sorry.
