Hello,
I’m using WSE to stream via rtsp. All works fine since the stream received by my WSE instance is correctly received and transcoded in HLS as I need.
Since I need to make it accessible to a possibly very large number of viewers, I configured WSE as a live origin application associating it in a cloudfront instance.
In this case I encountered some strange behaviors that are difficult to manage since they are difficult to predict.
The first strange behavior is the fact that if I start to send the rtsp stream to WSE and I start to request the stream through cloudfront, if I try to access too early the “playlist.m3u8” file and I get a 404 error because the playlist isn’t ready yet, I can’t access the playlist from the same client for a certain amount of time (30 seconds or more sometimes). My guess is that subsequent requests to the same playlist are cached for a certain amount of time, so I can’t properly implement a mechanism to play the stream as soon as it became ready.
My question is: is there a proper way to configure both Wowza and Cloudfront to prevent this behavior? I found a tutorial on the internet that suggests to set on cloudfront a minimum TTL of 0 for 404 and 500 errors. Is it the proper way to do this?
I noticed a similar behavior in case I use nDVR with cloudfront. If I request the stream from cloudfront too early, sometimes I get a 404 error not on the “playlist.m3u8” but on the first chunk of the playlist (i.e access “chunklist_b1396000_DVR.m3u8” gives me a 404 error) and the x-cache parameter in the response is
x-cache:Error from cloudfront
.
I guess the same as before, maybe I misconfigured my wowza or my cloudfront instances.
Is there a proper way to configure wowza?
Notice that if I request the stream without “?DVR” I never get a 404 error in requesting a chunck.
More in general. Is there exist a tutorial or a guide to deal with wowza + cloudfront cache?
I hope I have been clear.
Is a bit difficult situation to explain. Please ask me if something I wrote is not clear.
Thanks,
Simone
