Comments inline.

2011/9/27 Greg Logan <[email protected]>

> On 11-09-26 03:02 PM, Rubén Pérez wrote:
> > I have always found so absurd that the recordings alone have the
> > capability to report their state independently from the capture agent
> > they live in, as if they were separate entities and (which is worse) at
> > the same hierarchical level.
>
> Well... They are separate entities.  Imagine if the CA was unable to
> access the network for a day, but was otherwise happy and healthy.  It
> now has a backlog captures which it needs to update, along with the
> scheduled captures for a new day.  This requires that the recordings and
> the CA be (more or less) at the same hierarchical level - they're just
> on different branches of the tree.  Also, I would like to point out that
> the *CA* is what's sending the recording's state to the core, not the
> recording itself.  The recording has no brains, it's just a data structure.


Then why a simple data structure reports its state independently from: a)
other data structures of its kind; and b) the agent which is actually doing
the work? Why not link the recordings to the agent which is reporting their
state? I just don't see the benefit.

> Not to mention that a hierarchical structure is much better in terms of
> > assigning IDs, since two recordings may have the same ID without
> > colliding, if they belong to different agents.
>
> Why on earth would we want to assign the same ID to multiple captures?
> What good does this bring, considering the extensive refactoring costs
> on both the core and the CA?


Sorry, I was not very clear. I meant that you cannot assign IDs locally in
each capture agent because they're likely to collide with other agent's.
This may be a matter of opinion, but I found more understandable having "the
capture 5 in the agent X" than "the capture 43453453", and then having to
guess which agent it belongs to.


>
> > A xml-serialized Recording state is but a pair of fields indicating the
> > ID (perhaps the agent it belongs to) and the state. I don't see why an
> > agent state report cannot contain the state report for all the
> > recordings it contains. We are certainly sending much bigger xmls
> > through our endpoints without problems.
>
> True, but this upends what we've been talking about earlier in another
> thread:  If we want to drive adoption (both commercial service providers
> and end-users) then we need to maintain steady endpoints.
>

Yes, I know. But while we should keep that in mind, it's also true that some
endpoints are still changing to accomodate new needs or design changes.
I won't #propose a change in the reporting structure, of course, because I
don't think it's so urgent or will report so many benefits, but I do believe
it's more logical and well-organized, and avoids the agents report recording
states thousands of times when they could do it in a single call.

> Best regards
> > Rubén
> >
> > 2011/9/26 Judy Stern <[email protected] <mailto:[email protected]
> >>
> >
> >     Taking this to the level of the admin UI, there seem to be a number
> >     of use cases that call for users being able to see what capture
> >     agents were used/are being used/will be used for what recordings
> >     (and vice-versa).
> >     At this point (in the UI) one can see the capture agents listed for
> >     recordings in upcoming or capturing phases, but even there we can't
> >     yet sort the capture agents column or search for specific capture
> >     agents. And, of course, the Capture Agents tab should provide this
> >     information.
> >
> >     Related jira tickets: MH-2580, MH-7992, MH-2673, MH-3511
> >
> >     Judy
> >
> >
> >
> >     On Sep 26, 2011, at 10:39 AM, Greg Logan wrote:
> >
> >     > On 11-09-26 11:25 AM, Shane Phelan wrote:
> >     >> Is there somewhere on the core then that you can see a list of
> >     >> recordings for a particular agent?  And what the status is?
> >     >
> >     > Not on a per-agent basis, at least not on the core.  On the agent
> you
> >     > can see this information at /state/recordings, but the core does
> not
> >     > keep track of which recording comes from which agent.
> >     >
> >     >> I'm not sure if there is a problem at this point or not, but it is
> a
> >     >> concern.  Is there documentation somewhere of the capture agent
> >     workflow
> >     >> and the endpoints it uses?
> >     >
> >     > This is something that we need to work on.  I know as a developer
> >     which
> >     > endpoints are in use, but there is no documentation that I know of
> >     which
> >     > lists them explicitly.  In terms of the workflow this is somewhat
> >     > deliberately vague.  The core does not care how the recording is
> >     > captured, only that it is.  It presents a set of endpoints which
> any
> >     > agent can integrate with, but the actual sequence of actions
> >     inside the
> >     > CA is left up to the implementation.
> >     >
> >     > G
> >     >
> >     >> Thanks again!
> >     >>
> >     >> -Shane
> >     >>
> >     >> On Mon, Sep 26, 2011 at 12:56 PM, Greg Logan <[email protected]
> >     <mailto:[email protected]>
> >     >> <mailto:[email protected] <mailto:[email protected]>>> wrote:
> >     >>
> >     >>    On 11-09-26 10:43 AM, Shane Phelan wrote:
> >     >>> Thanks Greg... but doesn't that become a resource issue that
> >     every CA
> >     >>> continues to send upload_finished to the Core for every video
> >     that it
> >     >>> has captured.  Is there a mechanism that clears these out after
> some
> >     >>> amount of time?  Also what's the purpose it doesn't appear that
> >     >>    the Core
> >     >>> does anything with the message because the video has already been
> >     >>> ingested an processed.
> >     >>
> >     >>    The agent stops sending the state for a given recording once
> >     one of two
> >     >>    things happens:
> >     >>
> >     >>    1)  If the agent runs low on disk space it begins removing the
> >     oldest
> >     >>    ingested lectures, which also removes the state updates.
> >     >>    2)  If the recording is over a certain age (check
> >     >>
> >
>  
> felix/conf/services/org.opencastproject.capture.ConfigurationManager.properties
> >     >>    for this value) the CA deletes its local cached copy, which
> >     also removes
> >     >>    the state updates.
> >     >>
> >     >>    We keep the recordings around on the CA so that (worst case)
> >     you have
> >     >>    some leeway to recover lectures which somehow get lost or
> >     damaged on the
> >     >>    core.  Part of keeping them around, however, is keeping their
> >     state up
> >     >>    to date.  If this is causing issues please file a
> >     bug/community feature
> >     >>    request.
> >     >>
> >     >>    G
> >     >>
> >     >>> Thanks again,
> >     >>>
> >     >>> Shane
> >     >>>
> >     >>> On Mon, Sep 26, 2011 at 11:54 AM, Greg Logan
> >     <[email protected] <mailto:[email protected]>
> >     >>    <mailto:[email protected] <mailto:[email protected]>>
> >     >>> <mailto:[email protected] <mailto:[email protected]>
> >     <mailto:[email protected] <mailto:[email protected]>>>> wrote:
> >     >>>
> >     >>>    On 11-09-26 08:19 AM, Shane Phelan wrote:
> >     >>>> I've done a few test captures with my CA and I'm curious why it
> >     >>>> continues to send the following message over and over to the
> >     >>    Core.
> >     >>>     This
> >     >>>> doesn't seem necessary.  Is there a part of the process that is
> >     >>>    failing
> >     >>>> that makes the CA think the Core doesn't know about the
> >     >>    uploaded file?
> >     >>>
> >     >>>    Those are the normal state updates that the CA sends to the
> >     >>    core.  Your
> >     >>>    logging level is set to debug which is why you are seeing so
> >     >>    many of
> >     >>>    these.  You can either turn down the frequency of the updates
> (in
> >     >>>
> >     >>
> >
>  
> felix/conf/services/org.opencastproject.capture.impl.ConfigurationManager.properties)
> >     >>>    or change the logging level.
> >     >>>
> >     >>>    G
> >     >>>
> >     >>>> 10:14:06 DEBUG (AgentStateJob:158) - #443 - Sending
> >     >>    recording 51's
> >     >>>> state: upload_finished.
> >     >>>> 10:14:06 DEBUG (AgentStateJob:193) - #443 - State push
> >     >>>> org.opencastproject.capture.impl.jobs.AgentStateJob@6ae05134 to
> >     >>>> http://10.8.104.89:8080/capture-admin/recordings/51 was
> >     >>    successful.
> >     >>>> 10:14:06 DEBUG (AgentStateJob:158) - #443 - Sending
> >     >>    recording 25's
> >     >>>> state: upload_finished.
> >     >>>> 10:14:07 DEBUG (AgentStateJob:193) - #443 - State push
> >     >>>> org.opencastproject.capture.impl.jobs.AgentStateJob@6ae05134 to
> >     >>>> http://10.8.104.89:8080/capture-admin/recordings/25 was
> >     >>    successful.
> >     >>>> 10:14:07 DEBUG (AgentStateJob:158) - #443 - Sending
> >     >>    recording 101's
> >     >>>> state: upload_finished.
> >     >>>> 10:14:07 DEBUG (AgentStateJob:193) - #443 - State push
> >     >>>> org.opencastproject.capture.impl.jobs.AgentStateJob@6ae05134 to
> >     >>>> http://10.8.104.89:8080/capture-admin/recordings/101 was
> >     >>    successful.
> >     >>>>
> >     >>>>
> >     >>>>
> >     >>>> _______________________________________________
> >     >>>> Matterhorn-users mailing list
> >     >>>> [email protected]
> >     <mailto:[email protected]>
> >     >>    <mailto:[email protected]
> >     <mailto:[email protected]>>
> >     >>>    <mailto:[email protected]
> >     <mailto:[email protected]>
> >     >>    <mailto:[email protected]
> >     <mailto:[email protected]>>>
> >     >>>>
> >     >>
> http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
> >     >>>
> >     >>>
> >     >>>
> >     >>>    _______________________________________________
> >     >>>    Matterhorn-users mailing list
> >     >>>    [email protected]
> >     <mailto:[email protected]>
> >     >>    <mailto:[email protected]
> >     <mailto:[email protected]>>
> >     >>>    <mailto:[email protected]
> >     <mailto:[email protected]>
> >     >>    <mailto:[email protected]
> >     <mailto:[email protected]>>>
> >     >>>
> >      http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
> >     >>>
> >     >>>
> >     >>>
> >     >>>
> >     >>> _______________________________________________
> >     >>> Matterhorn-users mailing list
> >     >>> [email protected]
> >     <mailto:[email protected]>
> >     >>    <mailto:[email protected]
> >     <mailto:[email protected]>>
> >     >>>
> http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
> >     >>
> >     >>
> >     >>
> >     >>    _______________________________________________
> >     >>    Matterhorn-users mailing list
> >     >>    [email protected]
> >     <mailto:[email protected]>
> >     >>    <mailto:[email protected]
> >     <mailto:[email protected]>>
> >     >>
> http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
> >     >>
> >     >>
> >     >>
> >     >>
> >     >> _______________________________________________
> >     >> Matterhorn-users mailing list
> >     >> [email protected]
> >     <mailto:[email protected]>
> >     >>
> http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
> >     >
> >     >
> >     > _______________________________________________
> >     > Matterhorn-users mailing list
> >     > [email protected]
> >     <mailto:[email protected]>
> >     > http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
> >
> >     Judy Stern
> >     Educational Technology Services, UC Berkeley
> >     [email protected] <mailto:[email protected]>
> >
> >
> >
> >     _______________________________________________
> >     Matterhorn-users mailing list
> >     [email protected]
> >     <mailto:[email protected]>
> >     http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
> >
> >
> >
> >
> > _______________________________________________
> > Matterhorn-users mailing list
> > [email protected]
> > http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
>
>
>
> _______________________________________________
> Matterhorn-users mailing list
> [email protected]
> http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
>
>
_______________________________________________
Matterhorn-users mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn-users

Reply via email to