You are viewing an old version of this page. View the current version.

    Compare with Current View Page History

    « Previous Version 25 Next »

    This section discusses the Job pane, which is the most widely used part of the UI. From here, you can monitor and adjust jobs to get the most out of your farm.

    The Job List Panel

    The Job List Panel shows a column summary of the jobs submitted to the farm, and provides right-click actions to manage those jobs. There are a handful of columns displayed by default, though around 30 columns are available that can be specified in the Preferences.
     
    The "Job Counts" title will show the number of jobs in the jobList view, the number that have been retrieved from the supervisor, and the total number of jobs in the supervisor. If the "Displayed" number is less than the "Retrieved" number, you should check and see if you are filtering the retrieved jobs with one of the job state buttons or the "User" or "Search" filter controls. In the example image above, for example, the "Killed" jobs have been filtered out of the display.

    The "Get Next n Jobs" button will appear if you chose to limit the number of jobs retrieved from the supervisor at one time by setting the "Job Limit" value to a number greater than 0 in the WranglerView preferences:

    This feature is intended to be used to speed up the UI in environments that keep many thousands of jobs in Qube! at one time.

     

    Frames and Chunks Columns 
    Qube! jobs can contain agenda items that are chunks of frames, ie. a single agenda item can represent more than 1 frame. This occurs when the job was submitted with the range execution set to either "chunks of frames" or "partitions".

    The "Frames" column is shown by default, and will display the number of actual frames in the job, regardless of the number of chunks the frame range is divided into. This functionality is made possible by the presence of a "frameCount" value in the job's package. If this value is not present, the "Frames" column will revert to showing the number of chunks; this is the behavior for all version of the WranglerView prior to version 6.2.

    If the use of chunks and partitions is a common scenario in your farm, the user may want to enable the "Chunks" column in the GUI Preferences to display the chunk or partition count for jobs so the user may easily track the chunk or partition count for all jobs.

    Frame Average and ETA Columns 
    Columns for "Frame Average" (Avg) and "ETA" automatically calculate the estimated time it will take to complete running jobs based on previously rendered frames. It performs a simple calculation of determining the average time per frame, multiplying that by the number of frames remaining, and then dividing by the number of subjob processes running. It calculates this "on the fly" in WranglerView when the frame information is present, so selecting the desired job is necessary to have it perform the calculation. It will display the ETA if there are running subjob processes or leave it blank if there are not (since it can only move toward completion if there are running subjob processes).

    Frame Status Abbreviations
    These are used in various places in the UI. They are not too astonishing:

    • Pending
    • Blocked
    • Complete
    • Failed
    • Running
    • Killed
    • Bad Login
    • Waiting
    • Suspended 

    Frame/Process Status Tally Fields 
    There are individual columns that provide a running tally per job for pending, complete, running, blocked, killed, and failed frames or subjob processes.

    Frame/Process Summary Field 
    The "State Frames" column provides a summary of all frame or instance/subjob processes. It allows you to identify all the states for individual items within a job at a glance. The states are separated by a "/".

    Right-Click Popup Menu Actions
    The right-click menu allows the user to perform the following actions on the selected items:

    • Refresh Update the information on that Job from the Supervisor
    • Filter Dependencies Only display the jobs that are up or downstream dependencies of the selected jobs. This will fill in the Search Filter. Clear it to view all jobs again.

    • Retry... Retry the entire job. Also see the more commonly used "Resubmit" action below.
    • Retry Failed Frames... Retry only the frames listed as failed.
    • Interrupt... Kills the job, and then requeues it for another Worker (that is, it goes into the Pending state). The slider allows you to try to kill the N jobs which have been running for the shortest time, in order to throw away the least amount of work. This is the same as using qbinterrupt on the command line.
    • Block (+ Purge) Kill the job, and put it back in the queue in the Blocked state.
    • Block (+ Finish Current) Let the job finish, then kill the rest of its frames and put it back in the queue in the Blocked state.
    • Unblock Put a blocked job back in the active queue.
    • Shove... Tell the Supervisor to reevaluate this job. Useful if parameters have been changed with Prioritize or Modify (below).
    • Kill Kill a running or pending job. This job will not run again unless retried or resubmitted.
    • Remove... Remove the job from the Supervisor's MySQL database.

    • Prioritize... Adjust the job's priority on the fly. A lower number means a higher priority.
    • Modify... Adjust other tweakable job fields on the fly.
    • Resubmit... Submit a new job with the job properties of the existing job used as default.

    • View Output... View the output images in an external browser (if the Jobtype supports it)
    • Browse Output Dirs... Open a file browser to the specified files.

    • Execute in Local Shell... For commandline derived jobs, run the same command that the Workers call on the local machine. This is useful for troubleshooting.
    • Open Job Log Dir... Open the directory containing the logs for the selected job. This is useful for troubleshooting.
    • Zip and Open Job Log Dir... Zip up and open the directory containing the logs for the selected job. This is useful for troubleshooting and can be used to easily send the resulting logs out via email.

    • Copy Id's to clipboard...  Copy the job numbers to the computer's clipboard, for pasting into a document or email.
    • Save As... Save the selected job as a file (by default in an xml format). It can be submitted with the "Submit->Job from File" menu item.

     

    Job State Terminology

    Icon
    • When you retry a job, that means you attempt to assign it to new Workers as-is; none of the submission parameters are changed.
    • When you resubmit a job, you have the chance to change some or all of the submission parameters before it is assigned to a Worker or Workers.
    • When you kill a job, it is stopped immediately on all Workers. The job can be retried or resubmitted later, but you must choose to do that.
    • Blocking a job puts the instance or work item into a blocked state until something unblocks it. You can block a job immediately, in which case it is pulled from the Worker(s) and put into the blocked state, or you can ask that it wait for the current frame to finish first.
    • Interrupt removes a job (or several) from their current host(s), and then puts them back into the Pending state, where they can be picked up by another Worker. The slider allows you to interrupt N jobs, in which case it will sort all the current jobs by the amount of time they have been running, and interrupt the shortest-running ones. It does this to try to throw away the least amount of work.

    MetaJob Panel

    The "Meta Job Panel" is not displayed by default. It is displayed via the View menu, with the "View->Show/Hide: MetaJob Panel" menu item. There is also a hotkey, <Shift+Ctrl+M>, which will toggle it on and off.

    The MetaJob panel shows all jobs grouped under a single Process Group (pgrp) into a single item. This means that a complex process that may contain a dozen different jobs for rendering and compositing can now be viewed as a single item. It is a "Pull-Out" panel above the JobList Panel. It filters the JobList panel through the Search field.
    When a metajob is selected, it will automatically set the search filter to filter the Job List to only display jobs in that metajob.

    Most of the fields are the same as the JobList. Here are the fields unique to Meta Job Panel:

    • Pgrp: List of jobids under the Process Group (pgrp)
    • Jobs: Number of jobs in the Process Group (pgrp)
    • % Jobs Done: measure progress by number of jobs complete
    • % Tasks Done: measure progress by number of tasks complete for all jobs
    • Name: Name of the pgrp is determined from the job with the jobid matching that of the pgrp.


    A full set of actions can be reached in the MetaJob Panel by rt-clicking on a pgrp. Actions include:

    • Refresh: Refresh all jobs in the pgrp

    • Kill: Kill all jobs in the pgrp
    • Remove: Remove all jobs in the pgrp

    • Prioritize: Reprioritize all the jobs in the pgrp
    • Modify: Modify the properties of the jobs in the pgrp
    • Resubmit: Resubmit all jobs in the pgrp as a new pgrp

    • Deselect All: Deselects any selected pgrp and unfilters the JobList

    Frames/Work Sub-Panel

    The Agenda/Frames Pane lists the frames or items in the agenda that are to be processed.
    The right-click menu allows one to perform the following actions on the selected items:

    • Retry  Retry a frame. It will go back to the "pending"state and then will start when a Instance/Subjob process picks it up.
    • Block T ake a pending frame out of the active queue.
    • Unblock  Put a blocked frame back in the active queue.
    • Kill  Kill a running Instance/Subjob process.
    • Complete  Set a "pending" frame to the status of "complete".
    • Resubmit  Submit a new job with only the selected frames.
    • View Images  View the output images in an external browser (if the Jobtype supports it)
    • Browse Dirs  Open a file browser to the specified files.


    One can also select frames to filter the Stdout and Stderr panel text.

    Instances Sub-Panel

    The Instance Processes Pane lists the Instances (or processes) on the Qube Workers that are running the job.
    The right-click menu allows you to perform the following actions on the selected items:

    • Retry  Retry a killed or failed Instance process. It will go back to the "pending"state and then start when a Qube Worker becomes available.
    • Kill  Kill a running Instance process.
    • Migrate  Terminate the currently Instance process, place it back into the "pending" state for another available Worker, and prevent it from dispatching to the same host where it had been running.
    • Complete  Set a "pending" instance to the status of "complete".


    You can also select these Instances to filter the Stdout and Stderr panel text.

     

     

    • No labels