Jobs can be submitted from WranglerView using the items under the Submit menu. Selecting one will launch a modal submission dialog with the standard framework and job-specific parameters. Specify the fields and then press the "Submit" button to submit the job to the Qube! Supervisor.
Submissions for specific applications, such as Maya and Nuke, are explained in detail in their own section. Please refer to that documentation for submissions to individual applications. This page documents the complete set of parameters, which may not be needed or be exactly the same for individual applications.
There is a common set of buttons located at the bottom of the Submission Dialog.
- Set Defaults: Store as defaults in the User Preferences for that Jobtype, the values listed in the current submission dialog. The interface will use those values the next time the dialog is opened. This allows you to specify common fields like Priority or Executable that should always be the same value.
- Clear Defaults: Remove any stored defaults for that Jobtype submission dialog
- Expert Mode: Toggle button to display or hide export and non-default values from the submission dialog to reduce clutter when there are a lot of parameters. The current state of this toggle is stored in the Preferences Dialog.
- Save [Disk Icon]: Store the current properties of the job from the dialog as a file (by default an XML file). This can be used to submit through the WranglerView at a later time with "Submit->File…" or from the command line with the
qbsubcommand, or via the API.
- Cancel: Cancel the job submission and close the dialog.
- Submit: Submit the job with the specified parameters to the Qube! Supervisor.
Basic Parameters for all Jobs:
- Name: name used for the job, usually artist specified.
- Priority: priority of the job. A number between 1 and 9999. Lower numbers mean higher priority.
- CPUs: The number of processes (or subjobs/instances) to run the job with at the same time. When rendering, this equates to the number of frames being rendered at the same time.
Render Thread and Job Reservation Controls
This section is located near the top of the submission UI, and is collapsed when 'Expert Mode' is not selected. For applications that do not support setting the number of threads, this section is not visible at all.
Enabling 'Expert Mode' at the bottom of the UI will open this field up
Thread Control Behavior:
Checking Use All Cores for applications or renderers that support auto-detection of the number of cores installed on a worker host will set the renderer's appropriate control to enable this feature. In the case of a Maya job, it sets renderThreads=0; for the Mentalray renderer, it sets autoRenderThreads=True. It will try to "do the right thing" for each application where this control is visible.
Checking Render on all cores will also sets the job's reservation string to reserve all cores: you will set a "+" at the end of the host.processors=N+ . This means "only start on a worker with N free slots, but reserve them all, so that no other jobs will start on this worker while I'm running on it". The Min Free Slots value will affect the value of N, so setting it to 4 will say "start only on a worker with 4 free slots".
Render on all cores also disables the bottom-half of this section, Specific Thread Count and Slots=Threads, whose behavior is explained below.
Setting Specific Thread Count, on the other hand, does not necessarily reserve all the cores on the worker, it only sets the renderer's particular number of threads control, in Maya's case it's Render Threads:
But you will notice that the job's reservations have not changed, and is probably still at the default value of host.processors=1. This is to support workers which are configured for fewer slots than they have cores.
The Slots=Threads control will link the job reservations to the thread count, so a 4-threaded job will reserve 4 worker slots. It also disables the Render on all cores and Min Free Slots controls.
As you change the Specific Thread Count control the reservations value will automatically update.
Parameters for Cmdline Jobs:
- Command: The command to run on the Worker. Paths and syntax should be what the Worker's OS expects, not the submitting machine.
- Shell (Linux/OS X): Specify the shell to use when executing the command line on the Worker. Only visible in Expert Mode.
Parameters for Cmdrange Jobs:
- Command: Same as the Cmdline Job, though it will substitute the following strings based on the frame being executed for the given task.
- Range: The frame range to execute.
- Individual Frames
- Chunks with n Frames
- Split into n Partitions
- Ordering: Specify whether tasks should be executed in ascending, descending, or binary sort (first, last, middle, split the middle values, …) order.
Common Parameters for SimpleCmd Jobs:
- All the fields for either Cmdline or Cmdrange jobs (see above), plus
- Cmd Template: String used to construct the command line along with the rest of the job parameters. Python string representations are used, e.g. %(val)s to represent the string value from "val". If listed, %(argv)s places all optional arguments at that location instead of at the end. This constructs a command line that is then used by the Cmdline or Cmdrange jobtypes on the Workers. See above for the string replace values for CmdRange.
- Executable – path to the renderer or executable to run. Unless it is in the path on the Worker's environment, this needs to be set to an absolute path (to where the executable is located on the Worker)
Qube Worker Selection
Explicit list of Worker hostnames that will be allowed to run the job (comma-separated).
Explicit list of Worker groups that will be allowed to run the job (comma-separated). Groups identify machines through some attribute they have, eg, a GPU, an amount of memory, a license to run a particular application, etc. Jobs cannot migrate from one group to another. See worker_groups.
Explicit list of Worker hostnames that are not allowed run the job (comma-separated).
Explicit list of Worker groups that are not allowed to run the job (comma-separated).
Clusters are non-overlapping sets of machines. Your job will run at the given priority in the given cluster. If that cluster is full, the job can run in a different cluster, but at lower priority. Clustering
Order to select Workers for running the job (comma-separated) [+ means ascending, - means descending].
Host Order is a way of telling the job how to select/order workers
Worker properties needed to be met for job to run on that Worker (comma-separated, expression-based). Click 'Browse' to choose from a list of Host Order Options.
Requirements is a way to tell the workers that this job needs specific properties to be present in order to run. The drop-down menu allows a choice of OS:
You can also add any other Worker properties via plain text. Some examples:
With integer values, you can use any numerical relationships, e.g. =, <, >, <=, >=. This won't work for string values or floating point values. Multiple requirements can also be combined with AND and OR (the symbols && and || will also work).
The 'Only 1 of a "kind" of job' checkbox will restrict a Worker to running only one instance with a matching "kind" field (see below). The prime example is After Effects, which will only allow a single instance of AE on a machine. Using this checkbox and the "Kind" field, you can restrict a Worker to only one running copy of After Effects, while still leaving the Worker's other slots available for other "kinds" of jobs.
Worker resources to reserve when running job (comma-separated, expression-based).
Reservations is a way to tell the workers that this job will reserve the specific resources for this job.
Restrict job to run only on specified clusters ("||"-separated) [+ means all below, * means at that level]. Click 'Browse' to choose from a list of Restrictions Options.
Restrictions is a way to tell the workers that this job can only run on specific clusters. You can choose more than one cluster in the list.
Qube Advanced Job Control
List of submission flag strings (comma separated). Click 'Browse' to choose required job flags.
|See this page for a full explanation of flag meanings
Wait for specified jobs to complete before starting this job (comma-separated). Click 'Add' to create dependent jobs.
You can link jobs to each other in several ways:
The second menu chooses between "job" (the entire set of frames) and "work" (typically a frame). So to link frame 1 of one job to frame 1 of a second, job, you would choose "work" in this menu. If you want to wait for all the frames of one job to complete before starting a second, then choose "job". The other option, "subjob", refers to the instance of a job. This is much less common, but means that, for example, the instance of Maya that was running frames has completed.
For a complete description on how to define complex dependencies between jobs or frames, please refer to the Callbacks section of the Developers Guide.
Email (job complete)
Send email on job completion (success or failure). Sends mail to the designated user.
Email (failed frames)
Sends mail to the designated user if frames fail.
Set initial state of job to "blocked".
Redirect and consolidate the job stderr stream to the stdout stream. Enable this if you would like to combine your logs into one stream.
Optional label to identify the job. Must be unique within a Job Process Group. This is most useful for submitting sets of dependent jobs, where you don't know in advance the job IDs to depend on, but you do know the labels.
Arbitrary typing information that can be used to identify the job. It is commonly used to make sure only one of this "kind" of job runs on a worker at the same time by setting the job's requirements to include "not (job.kind in host.duty.kind)". See How to restrict a host to only one instance of a given kind of job, but still allow other jobs
Job Process Group for logically organizing dependent jobs. Defaults to the jobid. Combination of "label" and "Process Group" must be unique for a job. See Process group labels
Number of times to retry a failed frame/job instance. The default value of -1 means don't retry.
Retry Work Delay
Number of seconds between retries.
Kill the subjob process if running for the specified time (in seconds). Value of -1 means disabled. Use this if the acceptable instance/subjob spawn time is known.
Kill the agenda/frame if running for the specified time (in seconds). Value of -1 means disabled. Use this if you know how long frames should take, so that you can automatically kill those running long.
Qube Job Environment
- Cwd: (Current Working Directory) Explicitly specify the current working directory where the Worker should start the process
- Env vars: (Environment Variables) Explicitly set specific environment variables that should be used when running a job process on the Worker.
- Impersonate User: Submit job as specified user. Default is current user. Format: <optionaldomain>\<username> (advanced – requires Impersonate user permission)
Qube Output Parsing and Validation
- regex_highlights: Regular expression for highlighting information messages from stdout/stderr
- regex_errors: Regular expression for identifying fatal errors from stdout/stderr
- regex_outputPaths: Regular expression for identifying outputPaths of images from stdout/stderr
- regex_maxLines: Maximum number of lines to store for regex matched patterns for stdout/stderr
- validate_fileMinSize: Minimum size for identified outputPaths (in bytes). [0 disables test]
- generateMovie: add linked job to generate movie from output images. It will launch a second submission dialog to convert the images to a movie. This addresses the common action chain of rendering images and then converting those images into a Quicktime or similar movie. Note: This function will only work on jobtypes that retrieve the output image filenames into Qube (Maya, 3dsMax, Softimage, AfterEffects, etc).
- Login: login username for Shotgun
- Task: Shotgun task
- Project: Shotgun project
- Shot/Asset: Shotgun shot/asset
- Version Name: Shotgun version template. Performs variable substitution using python format %(key)s to get the value. The dict keys are prefixed with "job.", "package.", and "shotgun." respectively.
- Description: Shotgun description. Performs variable substitution using python format %(key)s to get the value. The dict keys are prefixed with "job.", "package.", and "shotgun." respectively.
- Account: Arbitrary accounting or project data (user-specified)
- Kind: Arbitrary typing information that can be used to identify the job - see How to restrict a host to only one instance of a given kind of job, but still allow other jobs
- Job Notes: Freeform text for making notes on this job
Require automatic drive mounts on worker.
Enable auto-wrangling for this job.
|Automatically convert paths on worker at runtime.
Normally instances are automatically completed by the system when a job runs out of available agenda items. Setting this flag disables that.
Normally, if a job is submitted with the number of instances greater than there are agenda items, Qube! automatically shrinks the number of instances to be equal to the number of agenda items. Setting this flag disables that.
(Deprecated in Qube6.5) Disable Windows' process management mechanism (called the "Job Object") that Qube! normally uses to manage job processes. Some applications already use it internally, and job objects don't nest well within other job objects, causing jobs to crash unexpectedly.
Submit job as an elite job, which will be started immediately regardless of how busy the farm is. Elite jobs are also protected from preemption. Must be admin.
Use environment variables set in the submission environment, when running the job on the workers.
(Deprecated in Qube6.5) Automatically expand job to use as many instances as there are agenda items (limited by the total job slots in the farm).
Wait for all instances to start before beginning work (useful for implementation of parallel jobs, such as satellite renders).
Run job on all candidate hosts, as filtered by other options (such as "hosts" or "groups").
Send e-mail when job is done.
When an agenda item (frame) fails but is retried automatically because the retrywork option is set, setting this flag causes the instances to be migrated to another worker host, preventing the frame from running on the same host.
|Prevent supervisor from applying supervisor_job_flags
Enable p-agenda for this job, so that some frames are processed at a higher priority.
Prevent job from being preempted.