Open jinsteros opened 1 year ago
Ok, found problem already.
To solve it I have to disable nginx healthchecks for upstreams because vod-module can produce only 502 errors unfortunately
And during this 502 on missing files, vod-module also produces this segfaults. As I found - all segfaults in logs are only from missing files requests. Example /Finland/dir2/1642/1642,144,240,360,480,720,p.mp4.urlset/master.m3u8 - if in the target dir we have only 144,240,360,480 mp4 files and no 720p.mp4, then accessing 1642,144,240,360,480,720,p.mp4.urlset/master.m3u8 manifest will cause a 502+segfault in system log in my case
Introducing 404 errors in vod-module can solve it, I think
Hello. Hope to find some help with my situation
I have 1 caching ssd server and 2 same storage servers with 4x10 TB HDD streaming HLS from small mp4 files (100-500 MB each one). Website is very old, I was streaming with the same config many years without changing anything and without any problems. But now after installing new storage servers with Debian 11 Bullseye I found strange 502 errors in my access log for 1-5% of traffic (both ts and m3u8 requests). In nginx error log on caching server this errors looks like
[error] 447386#447386: *352454516 upstream prematurely closed connection while reading upstream, request: "GET /Finland/59769/59769_720p.mp4/seg-39-v1-a1.ts HTTP/1.1", upstream: "https://x.x.x.x:443/Finland/59769/59769_720p.mp4/seg-39-v1-a1.ts" [...]
Then on storage servers (both of them) I found the reason as I think:
Many errors, every 10-20 seconds, randomly with ts or m3u8 requests, on both storage servers with same configs.
Nginx is configured like this:
At first I updated everything, but this can't solve the problem. So now I have:
Then I tried with core dump and gdb:
and no idea what is it and how to solve it. I would be very grateful for any tips