Just to say thanks guys - with workspace deployed and file.repo url defined all is working well!

D

On 02/16/2012 07:36 PM, David Horwitz wrote:
Aha thanks Ruben, that answers the question I was asking Andreas. I was using workspace-stubb so giving it a whirl with workspace ...

David

On 02/16/2012 06:27 PM, Rubén Pérez wrote:
David,

As you probably know, there are two ways of using the "workspace" module, needed by the admins and workers to share the files processed: if you compile your machine with -Pworkspace, then Matterhorn assumes that the files are local to the machine; in this case, local may also mean "mounted in the system via a distributed file system", such as NFS, but accessible as if it was local. If, instead, you use -Pworkspace-stub, then the file is requested via REST service to a machine which does have a local workspace, and is copied via HTTP. That's where the 2GB limitation comes, since (apparently) you can not send big files via HTTP, at least not directly. As most people does not use workspace-stub, this issue has not been fixed yet.

If you can put your files in a shared folder and mount it in your admins/workers, so that they can use "workspace" rather than "workspace-stub", then you're good to go. Otherwise, you can use your recently gained new committer status to put some pressure on this issue being fixed :P

Best regards
Rubén

2012/2/16 <[email protected] <mailto:[email protected]>>

    Sorry David, I can't really tell a lot, I'm not so deep into all
    this and rather the trial-and-error type ;)

    The other thing our admin did was to upgrade Java from version
    6U15 to 6U30. But that didn't really help here.

    What helped was the indicated line in the config on the *worker*
    node.

    Maybe just check if you get some log lines like
    2012-02-02 14:31:37  INFO (WorkspaceImpl:245) - Downloading
    http://server ....
    on your worker node, just before your mentioned error message starts.

    While it should not download the file (and hits the 2GB limit
    there) but rather get it from the shared storage.

    Andreas

    David Horwitz schrieb am Thu, 16 Feb 2012 betreff "Re:
    [Opencast...":

        Hmm ok that didn't work :-( Might be that I didn't quite
        understand what was needed. I now have:

            # The path to the repository of files used during media
            processing.
            
org.opencastproject.file.repo.path=${org.opencastproject.storage.dir}/files

            # The interval in milliseconds between two rounds of
            dispatching in the service registry. The default value is
            5s, and
            # a mimimum value of 1s is enforced due to performance
            reasons. Set to 0 to disable dispatching from this service
            # registry.
            #org.opencastproject.serviceregistry.dispatchinterval=5000

            # The base URL of the file server.  When using a shared
            filesystem between servers, set all servers to use the
            same URL.
            # If this is set to any value other than
            ${org.opencastproject.admin.ui.url}, the admin and file
            servers must use a
            # single sign on (SSO) solution such as CAS to avoid
            requiring users to perform multiple logins.
            
org.opencastproject.file.repo.url=${org.opencastproject.admin.ui.url}


        But it still fails :-( Did I get something wrong?

        David


        On 02/16/2012 04:35 PM, David Horwitz wrote:

            Thanks Andreas,

            In my test one the encoded file was 2.1G so this could be
            the cause. I've made the change in our dev environment
            and will report back once the test has run.

            D

            On 02/16/2012 04:27 PM, [email protected]
            <mailto:[email protected]> wrote:

                Hi David,

                we have been having the same issue for some videos.
                As you already said, this occurs only with some
                files, and the reason we found here for the failure
                is 2fold:

                1st the configuration admin-worker: as you already
                said it only happens there
                2nd the limit of 2GB for Jetty downloads: for some
                reason the worker *downloads* the prepared files, and
                if they are of a size > 2GB it will cut them at
                exactly 2GB wich causes the "moov atom not found" error.

                Our admin found this to fix the issue:
                in
                /opt/matterhorn/felix/conf/config.properties
                set
                
org.opencastproject.file.repo.url=${org.opencastproject.admin.ui.url}

                as suggested in http://opencast.jira.com/browse/MH-8032

                Hope this helps you also, regards, Andreas

                David Horwitz schrieb am Thu, 16 Feb 2012 betreff
                "[Opencast Matterhorn]...":

                    Hi All,

                    We're seeing failures on our production instance
                    in encoding certain files. The failure is
                    consistent and always happens at the encoding of
                    the presentation for the hold state preview. The
                    log message looks as follows:

                        2012-02-16 15:44:28 INFO
                        (AbstractCmdlineEncoderEngine:233) -
                        Executing encoding command: ffmpeg -strict
                        unofficial -i
                        
/data/matterhorn/workspace/mediapackage/eeaa4715-e6c5-444e-88eb-4c0835263a25/632f74bf-0dfe-4323-9a5a-ae6f44fbf619/Presentation.mp4
                        -r 10 -s 320x200 -vcodec flv -b 256000 -ar
                        11025
                        
/data/matterhorn/workspace/mediapackage/eeaa4715-e6c5-444e-88eb-4c0835263a25/632f74bf-0dfe-4323-9a5a-ae6f44fbf619/Presentation_b36b560a-3e58-4ee1-a17c-5d21f077b3ce-10fps-preview.flv
                        2012-02-16 15:44:30 INFO
                        (FFmpegEncoderEngine:174) - ffmpeg version
                        0.8.10, Copyright (c) 2000-2011 the FFmpeg
                        developers
                        2012-02-16 15:44:30 INFO
                        (FFmpegEncoderEngine:174) -
                        [mov,mp4,m4a,3gp,3g2,mj2 @ 0x13c5760] moov
                        atom not found
                        2012-02-16 15:44:30 INFO
                        (FFmpegEncoderEngine:174) -
                        
/data/matterhorn/workspace/mediapackage/eeaa4715-e6c5-444e-88eb-4c0835263a25/632f74bf-0dfe-4323-9a5a-ae6f44fbf619/Presentation.mp4:
                        Operation not permitted
                        2012-02-16 15:44:30 WARN
                        (AbstractCmdlineEncoderEngine:269) - Error
                        while encoding video track Presentation.mp4
                        using 'flash-preview.http': Encoder exited
                        abnormally with status 1
                        2012-02-16 15:44:30 WARN
                        (ComposerServiceImpl:249) - Error encoding
                        
http://mediadev.cet.uct.ac.za/files/mediapackage/eeaa4715-e6c5-444e-88eb-4c0835263a25/632f74bf-0dfe-4323-9a5a-ae6f44fbf619/presentation.mp4
                        and null



                    The error seems to happen when some or all of the
                    presentation is the epiphan capture card (seen
                    when no source was connected). I can reliably
                    reproduce this with certain recordings on both
                    our production and development environments
                    (separate admin/worker) but not my local build
                    (all in one). This, with the error "Operation not
                    permitted" suggests some kind of workflow/file
                    repository error - the ffmpeg docs lean towards
                    this being 1 of the files doesn't exist.  However
                    this doesn't explain its reproducibility with
                    certain packages and not others.

                    So far I have verified:

                    1) The profile creates a valid working copy (i.e
                    ffmpeg isn't encoding to to 0b file)
                    2) the file doesn't exist at:
                    
/data/matterhorn/workspace/mediapackage/eeaa4715-e6c5-444e-88eb-4c0835263a25/632f74bf-0dfe-4323-9a5a-ae6f44fbf619/Presentation.mp4
                    But I don't know if one expects it to be cleaned
                    up after the failure ...


                    Any help in solving this is apreciated

                    David



                    _______________________________________________
                    Matterhorn mailing list
                    [email protected]
                    <mailto:[email protected]>
                    http://lists.opencastproject.org/mailman/listinfo/matterhorn


                    To unsubscribe please email
                    [email protected]
                    <mailto:[email protected]>
                    _______________________________________________


                -----------------------
                [email protected] <mailto:[email protected]>
                01/58801 DW 41523
                mobil: 0664/60 588 4523
                TU Wien
                DVR-Nummer: 0005886
                -----------------------
                _______________________________________________
                Matterhorn mailing list
                [email protected]
                <mailto:[email protected]>
                http://lists.opencastproject.org/mailman/listinfo/matterhorn


                To unsubscribe please email
                [email protected]
                <mailto:[email protected]>
                _______________________________________________


            _______________________________________________
            Matterhorn mailing list
            [email protected]
            <mailto:[email protected]>
            http://lists.opencastproject.org/mailman/listinfo/matterhorn


            To unsubscribe please email
            [email protected]
            <mailto:[email protected]>
            _______________________________________________


        _______________________________________________
        Matterhorn mailing list
        [email protected]
        <mailto:[email protected]>
        http://lists.opencastproject.org/mailman/listinfo/matterhorn


        To unsubscribe please email
        [email protected]
        <mailto:[email protected]>
        _______________________________________________


    -----------------------
    [email protected] <mailto:[email protected]>
    01/58801 DW 41523
    mobil: 0664/60 588 4523
    TU Wien
    DVR-Nummer: 0005886
    -----------------------
    _______________________________________________
    Matterhorn mailing list
    [email protected]
    <mailto:[email protected]>
    http://lists.opencastproject.org/mailman/listinfo/matterhorn


    To unsubscribe please email
    [email protected]
    <mailto:[email protected]>
    _______________________________________________




_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn


To unsubscribe please email
[email protected]
_______________________________________________


_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn


To unsubscribe please email
[email protected]
_______________________________________________
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn


To unsubscribe please email
[email protected]
_______________________________________________

Reply via email to