Differences

This shows you the differences between two versions of the page.

--- software:relive [2016/10/02 17:42] – andi
+++ software:relive [2024/01/03 16:49] (current) – [Relive integration in media.ccc.de (voctoweb)] andi
@@ Line 1: / Line 1: @@
 = HLS Relive
-Documentation for the relive system is contained within its git repo: ''git@c3voc.de:hls-relive'' -> README.md
+Documentation for the relive system is contained within its git repo ''git@c3voc.de:hls-relive'' mirrored at https://github.com/voc/hls-relive -> README.md
-It's currently deployed at live.ber:/home/relive/git/
+It's currently deployed at live.alb:/home/relive/git/
-== README
-<code>
+== Setting up Relive in the VOC setup
-HLS ReLive
-==========
-This set of tools records portions of an HLS stream into separate
+Login to server
-playlists for timeshifting and to bridge the time until the actual
+  ssh relive@relive.c3voc.de
-recordings are available.
+  cd git
+  screen -dR
-Components
+  - Create an appropriate config for your event in `live.alb:/home/relive/git/configs`
-----------
+    - either by using `./gen_event_config.py <conference slug>`, or
+    - manually: copy `cfg.example`, look at other files in that directory, or in the `old/` subdirectory for inspiration. The filename will be used as the conference name from here on.
+  - After a minute, a cronjob will run and download the schedule, etc. (If you are impatient, you can run `/home/relive/git/scripts/foreach-project.sh /home/relive/git/scripts/get-fahrplan.sh` manually)
+  - Start the scheduler (run as user relive, inside of a screen/tmux session): `./launcher.sh <conference slug>`
+  - In the streaming-website configuration, set ''RELIVE_JSON'' to point to ''%%http://relive.c3voc.de/relive/<conference slug>/index.json%%''
-The ReLive system consists of several tools and cron jobs, some of which
+Note that the ''index.json'' file is also only regenerated periodically, so it might be a minute or two before that URL actually works.
-can be used independently from the rest.
-### record.pl
-`record.pl` records HLS streams that are being generated locally by
+== Relive integration in media.ccc.de (voctoweb)
-watching the corresponding directory with inotify. It cannot record
-streams via HTTP, nor will it perform any magic (like trying to figure
-out at what position in the source m3u8 it stopped recording) when being
-restarted: It will simply add a discontinuity header to the output m3u8
-and resume adding segments. It turns out that this makes things much
-simpler and is enough for our use case.
-Usage: `record.pl in-directory in-m3u8 out-directory`
+When an event in voctoweb has no recordings, voctoweb will try to find the corresponding relive recording. This only works if the conference acronym matches the acronym used in the relive system.
-The m3u8 in the output directory will be named `index.m3u8`.
+To use this feature, first create the conference in voctoweb and in c3tracker. In theory there are multiple variants, but this one is the one we currently use:
-### scheduler.pl
+* To enable this workflow for you conference, enable the "releasing-relive-voctoweb" worker group which is handled by a modified version of CRS worker `script-A` living in a screen (user voc).
+* [[https://github.com/voc/voctopublish/blob/feature/relive/voctopublish/create-event-by-ticket.py|create-event-by-ticket.py]] is able to create an event in voctoweb, based on the CRS Ticket ID.
+* This script can be called during any state transition, we currently use `scheduled` -> `recording` of the recording ticket
+  * TODO: When we use this step, we should rename or migrate the deployment of of this script from releasing to storage, as it confuses tracker operators when the releasing host does the recording step.
-`scheduler.pl` is responsible for starting and stopping recording
+If you need re-create the session, run following commands:
-processes as dictated by the conference schedule. See below for
-configuration details.
-When sent a `SIGHUP`, it will reload its schedule, spawning or killing
-recording subprocesses as necessary. Running recording subprocesses
-which are still valid w.r.t. the new schedule remain untouched.
-When sent a `SIGINT` or `SIGTERM` it shuts down all recording
+```
-subprocesses and then terminates itself.
+ssh voc@releasing.c3voc.de
+screen -x
+cd /opt/crs/tools/tracker3.0
+source tracker-profile-relive.sh
+export LOCALFILE=tracker-profile-relive.sh
+./bin/crs_run script-A-recording+relive-scheduler.pl
+```
-Recording subprocesses are started via `wrapper.sh` which takes care of
+== README
-setting up their environment (creating one directory per schedule event
-to store the recording, amongst other things) and starting them.
-### genpage.pl
-`genpage.pl` looks through the directory with all the recordings,
-cross-referencing them with the schedule and media.ccc.de, to find out
-whether a proper release has happened yet. All this data is collected
-and written into a file called `index.json` in the top-level recording
-directory, which can then be used by the streaming frontend.
-A recording can be in one of four states:
- - not_running: Transitory state where the recording directory has been
-   created, but doesn't contain any data yet.
- - live: There is data present, but the playlist is not finished yet.
- - recorded: There is data present and the playlist is finished.
- - released: A recording of this talk has been found on media.ccc.de
-A recording can move from the *recorded* state back to *live* if the
-recording is restarted for some reason. Other than that, recordings
-progress monotonically from *not_running* to *released*.
-Starting from the *live* state, a thumbnail is generated. Once a
-recording is *recorded*, the HLS files get remuxed into a faststarted
-mp4 for easy download.
-The `index.json` file contains an array of objects. A fully populated
-object looks like this:
-    {
-       "thumbnail" : "//live.dus.c3voc.de/releases/relive/1549/thumb.jpg",
-       "status" : "released",
-       "duration" : 5399,
-       "room" : "HS 7",
-       "playlist" : "//live.dus.c3voc.de/releases/relive/1549/index.m3u8",
-       "id" : "1549",
-       "title" : "Btrfs – Das Dateisystem der Zukunft?",
-       "release_url" : "http://media.ccc.de/browse/conferences/froscon/2015/froscon2015-1549-btrfs_das_dateisystem_der_zukunft.html"
-    }
-### check_released.pl
-`check_released.pl` is a helper used by the `get-releases.sh` script. It
-ensures that only events which have at least one recording on media.ccc.de are
-marked as *released*.
-### Cron jobs
-There are three cron jobs:
-  - calling get-releases.sh to download and cache the list of recordings
-    already released on media.ccc.de
-  - calling get-fahrplan.sh to download the schedule
-  - calling genpage.pl to update `index.json` etc.
-Configuration
--------------
-All scripts take the union of `global_config` in the root of the git
-repository as their configuration. This configuration is extended by a
-project-specific configuration in the `configs` subdirectory. See the
-`.example` files in the respective directories. These files also contain
-comments explaining the various options.
-Setting up for a new conference
--------------------------------
-The following steps are necessary to set up ReLive for a new conference:
-  - update the configuration file appropriately
-  - launch the recording scheduler: `./launcher.sh projectname`
-    You'll probably want to do that in a screen session.
-</code>
-== Anleitung (in progress)
-  ssh voc@live.ber.c3voc.de
+Please read the [[https://github.com/voc/hls-relive/blob/master/README.md|README]] in the repository, it is being kept up to date with regard to the general behavior of the relive system.
-  sudo su relive
-  screen -x
-  # weiter siehe unten
-  * ''[9|13:20] <florolf> bjoern: ich hab relive gerade angeworfen. das beschraenkt sich auf config bauen und dann scheduler.pl in 'nem screen starten (RELIVE_PROJECT=configname perl scheduler.pl ../data/conferencename/schedule.xml)''
-  * ''[9|13:26] <florolf> bjoern: die streaming-website hat einen cronjob, der sich das relive-manifest runterzieht. das iteriert aber automatisch ueber alle existierenden konferenzen. pro konferenz muss da nur die download.sh richtig angelegt werden (siehe existierende configs auf lb1)''
+== FAQ
+The `/video/relive/<conference>/<id>/index.m3u8` is empty/broken, how can I regenerate the file?
+  * simply rename or remove the file, it will be generated in the next run (or trigger it manually)