Random broken video from yt2009sabr.js

[sh20000sh]

It usually happened from longer(like hours) videos with probability with not just to be ignored. Sometimes video are just stopped from certain point and only audio is being played. Tried to play video in question on instance locally, those are processed like that. Tried to download same video it became playable so it isn't video specific issue. I think some fragments cause later part of video corrupt for some reason.

[ftde0]

try changing the video quality if that happens, or skip around 30 seconds and see if it plays normally again, any video+quality combination you can get this consistently on?

[sh20000sh]

Usually watching with 720p. Skipping seems didn't work as I remember.
At first, I thought this was video-specific but completed video has small size and being not deleted by auto-maintain and making it looks like issue happening consistently.

Last case I got was from v=UfMyvDI_uxc around its 35 min point but it won't help much I think.

[sh20000sh]

I thought this was video-specific but completed video has small size

Checking audio and video length before merging might work I guess? Such like if they are not match, make it download video again.

[sh20000sh]

Random guess, it might be affected by certain time of day like guess from #287 did. I experienced two times in row from different point of same video, as I retried download right after it broken after like five minutes.

[ftde0]

Sometimes video are just stopped from certain point and only audio is being played

how long for? if it's just a few seconds before audio stops as well, that would mean that sabr stops sending new segments for a specific time offset request (that is known to happen rarely).

just for the record, this does happen on other browsers as well, right? any errors in console?

[sh20000sh]

how long for? if it's just a few seconds before audio stops as well, that would mean that sabr stops sending new segments for a specific time offset request (that is known to happen rarely).

As I saw, more than five minutes. When this happened, stored video in assets folder are also broken like that so I think checking other browser is not necessary. In most cases, delete that certain video and redownloading fixed this issue. Console show nothing in prod mode.

[ftde0]

wait, just to clarify, we're talking about sabr downloading or sabr playback directly in the browser (without downloads)?

[sh20000sh]

Ah yes. I was talking about downloader. Not about direct playback.

[ftde0]

hm, as with #287 you mentioned, i can't replicate this issue on my end (probably region based again), i'll add a temporary config option next update to add more extensive logging for sabr downloader to see if there is a simple fix

Checking audio and video length before merging might work I guess? Such like if they are not match, make it download video again.

i'd rather not do that if possible for the obvious reason of having to redownload a file entirely

[sh20000sh]

Tried that config, I saw one got deadlock around somewhere middle point (17:05 of 27:20 long video, needed to be interrupted to get thing back to working). What I experienced for this issue was not deadlock so not actually sure about this.

[ftde0]

the config option is only meant to log more info and doesn't fix the issue, if you get broken files at a specified offset then please share the log from around the offset the breakage happens (+ a few more before and after) so i can try to do some changes

[sh20000sh]

That deadlock seems not the case, just something messed up during log printing. I got actual issue now. Unfortunately, I found nothing special from that log. This broken around 2:35 of video and log around those were this:

[sabr] got response for offset 2:00 - 3 ps
[sabr] got response for offset 2:05 - 4 ps
[sabr] got response for offset 2:10 - 3 ps
[sabr] got response for offset 2:15 - 4 ps
[sabr] got response for offset 2:20 - 3 ps
[sabr] got response for offset 2:25 - 4 ps
[sabr] got response for offset 2:30 - 3 ps
[sabr] got response for offset 2:35 - 3 ps
[sabr] got response for offset 2:40 - 1 ps
[sabr] got response for offset 2:45 - 2 ps
[sabr] got response for offset 2:50 - 2 ps
[sabr] got response for offset 2:55 - 2 ps
[sabr] got response for offset 3:00 - 1 ps

[ftde0]

yeah, nothing standing out for me either here, can you reproduce the same deadlocks/issues on flash sabr? it also requests offsets like that so the issue should show here if it's network related

[sh20000sh]

This thing happens randomly and happen like once on more than a day, so I'm not sure I'm going to try it. As I saw, main point of issue was processed video on instance so direct playback won't replicate those exact issue. It would even play from middle of video so worst case would be just five sec long pause if it even happens.

After seeing that log, I think this happens on next step of downloading. Are there any possibilities that those bytes declare end of video unintentionally?