HLS cleanup not working

[hpms1972]

Hi,

i have the problem that hls_cleanup is not working at all, nginx does not delete anything, the path with the fragments is getting bigger and bigger until the is no space left on the disk
Here is my config:

        hls on;
        hls_path /tmp/hls;
        hls_fragment 10s;
        hls_playlist_length 10s;
        hls_cleanup on;

I used different settings for hls_fragment and hls_playlist_length but that does not change anything. Even with the above settings in a short time i get easily over 100 fragments without one of them is getting deleted.

Nginx is running on an RaspberryPi with an Arch-Linux.
Any ideas?

Thanks,
hpms1972

[arut]

what's in error.log?

[hpms1972]

Unfortunately the error.log is empty.

[arut]

Then full debug log
However most probably it's configuration issue. Please check your configuration once again.

[towolf]

He got the config from me: It's posted here (sorry for the bad colors)

We checked the GOP period also. The cam sends a keyframe every 250 frames at 25 fps so that should divide into neat 10s chunks.

[hpms1972]

So now i made a full debug log for a time period of approx. 12 min. (started at 18:34, ended at 18:46), 71 fagments where made in this time with the above posted settings, with 'grep cleanup' i get the following output:

[root@alarmpi tmp]# grep cleanup nginx.error.log
1970/01/01 00:00:18 [debug] 234#0: hls: cleanup path='/tmp/hls' playlen=10000
2013/10/07 18:46:13 [debug] 233#0: run cleanup: 017D6F50
2013/10/07 18:46:13 [debug] 233#0: cleanup resolver
2013/10/07 18:46:13 [debug] 232#0: run cleanup: 017D6F50
2013/10/07 18:46:13 [debug] 232#0: cleanup resolver

Are there more interesting parts you may want to see?

Thanks,
hpms1972

[towolf]

Here's the full log he sent me in email (3MB gzipped)

http://ftp.tuebingen.mpg.de/pub/kyb/towolf/raspberry-camera-streaming/resources/nginx.error.log.gz

[towolf]

Err, why does it say: playlen=10000?

note to self: milliseconds

[towolf]

Does it matter that on startup the clock is still set to 1970-01-01?

It might be that NTP network time comes later. It is visible in the log.

[hpms1972]

Indeed the time seems to be the problem.
When nginx starts after time-sync all works fine.

Many thanks to towolf and arut.

[jhanvi22]

I have the same issue.
I have set hls_cleanup on; in nginx conf file.
Still it is not removing old hls segments. How did you fix it?

[Krautmaster]

why closed?