Recherche avancée

Sur d’autres sites (5356)

• What are the gotchas of using statically linked libraries in serverless platforms such as Google Cloud Functions ?

  5 septembre 2017, par Dzh

  Libraries such as ffmpeg-static upload statically linked binaries onto container.

  I wonder what are the drawbacks of using this approach ?

  Does the library size counts against your memory use (it’s billed by GCloud) ?

  Does it slow down the container ? Perhaps some future-proofing issues ?

  Edit : Found something of a related (I wanted to setup OpenCV) on AWS blog. It doesn’t explain drawbacks, just shows how to do it exactly.

• In my django app, celery task converts uploaded video w/ ffmpeg, but converted video won't save to s3 ?

  29 janvier 2013, par GetItDone

  I use Heroku to host my website, and Amazon s3 to store my static and media files. I have a celery task that converts the video file to flv, but the flv doesn't store anywhere. There isn't any error, just there is no file uploaded to my s3 bucket. How can I force the file to save to my s3 bucket after the conversion ? I'm still pretty new web development in general, and I have been stuck trying to get my video conversion working properly for weeks. To be honest, I'm not even sure that I'm doing the right thing with my task. Here is the code in my celery task :

  @task(name='celeryfiles.tasks.convert_flv')
  
  def convert_flv(video_id):
  
      video = VideoUpload.objects.get(pk=video_id)
  
      filename = video.video_upload
  
      sourcefile = "%s%s" % (settings.MEDIA_URL, filename)
  
      vidfilename = "%s.flv" % video.id
  
      targetfile = "%svideos/flv/%s" % (settings.MEDIA_URL, vidfilename)
  
      ffmpeg = "ffmpeg -i %s -ar 22050 -f flv -s 320x240 %s" % (sourcefile, targetfile)
  
      #The next lines are code that I couldn't get to work in place of the line above, and are left commented out.
  
      #I am open to suggestions or alternatives for this also.
  
      #ffmpeg = "ffmpeg -i %s -acodec libmp3lame -ar 22050 -f flv -s 320x240 %s" % (sourcefile, targetfile)
  
      #ffmpeg = "ffmpeg -i %s -acodec mp3 -ar 22050 -f flv -s 320x240 %s" % (sourcefile, targetfile)
  
      try:
  
          ffmpegresult = commands.getoutput(ffmpeg)
  
          print "---------------FFMPEG---------------"
  
          print "FFMPEGRESULT: %s" % ffmpegresult
  
      except Exception as e:
  
          ffmpegresult = None
  
          print("Failed to convert video file %s to %s" % (sourcefile, targetfile))
  
          print(traceback.format_exc())
  
      video.flvfilename = vidfilename
  
      video.save()

  My view :

  def upload_video(request):
  
      if request.method == 'POST':
  
          form = VideoUploadForm(request.POST, request.FILES)
  
          if form.is_valid():
  
              video_upload=form.save()
  
              video_id=video_upload.id
  
              video_conversion = convert_flv.delay(video_id)
  
              return HttpResponseRedirect('/current_classes/')
  
      else:
  
      ...

  Any advice, insight, or ideas in general would be greatly appreciated. Obviously I am missing something, but I can't figure out what. I have been stuck with different aspects of getting my video conversion to work with ffmpeg using a celery task for weeks. Thanks in advance.

• How to build a daemon to encode video files on S3 ?

  4 avril 2013, par Yuval Cohen

  I am interested in running a daemon to go over user uploaded video files and encode them in an optimal format (and add some watermarks).

  I was considering services such as Zencoder, Encoding.com, Amazon's encoding service but some lack overlaying capabilities and some are just too expensive for our (big) volumes.

  I want to build a daemon that encodes videos that are located on S3 once users upload them.

  The solution I thought of would be Python Heroku servers using Celery for a task queue to keep track of the encoded files and ffmpeg to do the actual work. However, I ran into troubles compiling ffmpeg for Heroku (with libass support, so the basic ffmpeg bins aren't enough).

  What approach/technology stack would you consider for this mini-project ?

  Thanks !
  Yuval