Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Nicolas Menciere
    @koxon
    one for validation
    the other for transcoding
    this way you have 2 clusters that you can scale independently
    in order to test this, you will have to put yourself in situation
    keep a small cluster and send it a lot to do
    and see the alerts you receive over time
    jessica88888
    @jessica88888

    Dear Nicolas,

    Thank you so much for your encouragement.

    When I try to upload fake mp4 file, it passes through the Validate stage in State Machine and proceeded to the Transcode stage. May I know is the validation stage working correctly and is this step necessary?

    Sometimes, the transcode step failed in Step Function and returned:

    "error": "States.Timeout"

    May I know is this caused by termination of the cluster when the task still running, if yes, how can we solve this issue?

    Thanks Nicolas, you’re really kind. :smile: :thumbsup:

    Best regards,
    Jessica

    Nicolas Menciere
    @koxon
    What kind of fake mp4 you try to transcode ?
    Validate steps is supposed to return you the mime type and the ffprobe output
    if you dont have that the transcoding step should fail
    you may be able to add a stop in your state machine that checks the output of the Validate task
    if you shut down a worker while it's processing a task in a state machine then yes
    the State will timeout
    jessica88888
    @jessica88888

    Hi Nicolas,

    I’ve created an invalid MP4 file by simply renaming a text file with an extension of MP4 to see whether it goes through the validation steps. So, the validation step returned these metadata.

    "input_metadata": { "mime": "text/plain", "type": "text" },

    May I know what’s the difference between these two assets:
    https://github.com/bfansports/CloudTranscode/blob/master/state_machines/SATranscodeAssets.json
    https://github.com/bfansports/CloudTranscode/blob/master/state_machines/SAValidateTranscodeAssets.json

    Can I use SATranscodeAssets.json which doesn’t have the validation step?
    Thanks Nicolas.
    Best regards,
    Jessica

    Nicolas Menciere
    @koxon
    yes you can use the transcode State machine only
    in SFN console, check the jobs that you ran
    you can check the output of each step
    look at what the Validate step output
    and see what is says with your bad file
    look also at the transcode step
    jessica88888
    @jessica88888
    Hi Nicolas,
    Thanks for your explanation. I’ve tried to transcode without validation step, everything goes nicely.
    Currently, I’m still working on auto-scaling part.
    Many thanks for all of your support. :smiley:
    Best regards,
    Jessica
    Nicolas Menciere
    @koxon
    the auto scaling is quite painful indeed
    not sure how you can avoid killing a server that is running a task
    jessica88888
    @jessica88888
    Hi Nicolas,
    Yes, I agree with you. These few days, I’m trying to scale the task and instance up and down with different alarms. However, tasks that are running will still accidentally killed even with the lifestyle hook for instance.
    Hoping to find a solution that can allow the task to run till complete.
    Thanks Nicolas.
    Best regards,
    Jessica
    Nicolas Menciere
    @koxon
    maybe the scaling awareness should be built in the application itself
    Alarms could trigger a Lambda function
    this function could then send messages to the workers
    and order some workers to stop listening for new tasks
    but it sounds quite hard thoughg
    the workers would need to know which machine they running on, etc
    jessica88888
    @jessica88888
    Hi Nicolas,
    I will keep on trying on auto-scaling part, will update you if I found any solution on this, thank you so much.
    Best regards,
    Jessica
    jessica88888
    @jessica88888
    Dear Nicolas,
    May I know is there any way to set the permission of transcoded files?
    Thank you.
    Best regards,
    Jessica
    Nicolas Menciere
    @koxon
    Hi Jessica
    What you can do is update the permissions in the ClientInterface Class
    if you implemented one
    when transcoding is done
    you can call S3 and update the permissions of you alternative files
    shahzaibcb
    @shahzaibcb
    Hi Nicolas thanks for the quick reply on github.
    Nicolas Menciere
    @koxon
    np! Anytime
    jessica88888
    @jessica88888
    Dear Nicolas,
    Thank you so much for your supports. After few weeks of testing, I’ve got everything running smoothly on AWS. Really appreciate for your efforts. Thanks once again.
    Best regards,
    Jessica
    Dbinutu
    @Dbinutu
    Hi Nicolas, I also want to say a big thank you. Cheers Bro!
    Nicolas Menciere
    @koxon
    hey guys
    no problem, you're welcome
    @jessica88888 you managed to make auto scaling work ?
    jessica88888
    @jessica88888
    Hi Nicolas,
    I managed to auto-scale the instances and tasks. However, I’m not using lifecycle hook, as I couldn’t make it to work the way it should. The only thing I can do to reduce the task being killed is to increase the memory so that the job can run as fast as possible. Other than that, I also add more retries in step function. In case that the task was killed due to scale down of instance, it will still restart another task.
    Thanks Nicolas.
    Best regards,
    Jessica
    Nicolas Menciere
    @koxon
    Nicely done @jessica88888 you're doing better than me!
    best
    jessica88888
    @jessica88888
    Hi Nicolas,
    Thank you so much. Really appreciate for your supports and the great CloudTranscode.
    Thanks once again.
    Best regards,
    Jessica
    Alouane Nour-Eddine
    @alouane
    Hi @koxon
    I have questions regrading this great project