The VRED Cluster Manager organizes the shared use of a central cluster by multiple users. It provides you the ability to book certain time slots and resources for cluster raytracing on a one-time or periodic basis. Resources are booked for a user and each user must be registered, by an administrator, to use the system. You can see your reservations and the remaining available resources. Administrators manage all reservations and may stop sessions as necessary. A central server must be selected to work as the management server. The VREDClusterService does not require a license, but must be installed on the PC in use. Administrators can upload cluster configurations. The VRED Cluster Module should be used to create these configurations.
If Substance variant states won't render in the ClusterManager, see Unable to Render Substance Variant States on a Cluster Queue.
Click Rendering > Cluster Manager.
If the management PC has been configured, you can log in with the cluster manager. A user name and password are required.
If the cluster service is installed on the management PC, each PC in the network can be used as a management node. When you log in for the first time, a dialog informs you that the cluster manager is initializing with the given user as an administrator.
You can also execute the initialization from the command line. The command VREDNodePoolService -i, resets the database and asks for the name and password of an administrator.
All settings and booked sessions are stored in a database file. The location of this file depends on the operating system.
Windows, if VREDClusterService is running as windows service:
C:\\Windows\\System32\\config\\systemprofile\\AppData\\Roaming\\VREDClusterManager\\nodePool.db
Windows: C:\\users\\USER\\AppData\\Roaming\\VREDClusterManager\\nodePool.db
Linux: ~USER/.VREDClusterManager/nodePool.db
You can control the location of these files with the environment variable VRED_CLUSTERMANAGER_PATH. For a running system, you can view the info page to see the location of the database file.
If the cluster service is running at the first login, users should start registering with the system. The most convenient way to use the system is to use the windows login names and register the users without a password. In this case, users are automatically logged in. If circumstances required the use of passwords, a login screen is displayed when the Cluster Manager module is opened.
If users are added, one or more clusters must be configured. Create a cluster configuration with the cluster module. This configuration should include all available resources.
You can use localhost in configurations. Compute nodes named localhost, 127.0.0.1, or ::1 is ignored during scheduling.
If configurations contain display cluster definitions, then all display nodes must be available to start the configuration. If not all display nodes are available, the resource overview shows 0% available resources.
The resource overview shows available resources for an entire week. Available resources are shown as percentages and the actual number of available compute nodes. If all resources are available, the time slot is green. If no resources are available, the slot is red. If multiple clusters are available, the available resources can be shown for each cluster configuration.
To reserve nodes, select a time from the resource dialog and select Book nodes from the context menu. The booking editor is displayed.
To reserve nodes on a repeating basis, select the Type Periodic in the Bookings tab. Select a Start and End date for the rendering. Select Add Time Range to choose days and times for rendering to occur and save.
Sessions that have been successfully booked can be started by selecting the booking, then the Start button. If the start time has already been reached, the cluster session starts immediately. If the start time has not been reached, a countdown is displayed until the session begins.
The administrator determines what happens at the end of a session. A warning dialog can be shown before the end, the session can end automatically, or a dialog box can be shown that the session has ended.
There are different ways to stop a running session. You can stop running sessions by selecting the stop button in the Cluster Manager module or by selecting Stop in the cluster module. If a session was started on the local PC, an orange arrow is shown. If the session was started on another PC in the network, a blue arrow is shown. With this method, you can stop your session from any PC in the network. Administrators can use this function to stop sessions started by other users. If an administrator stops a session from other users, the session is marked as aborted.
The cluster manager can be used to do offline rendering. If a job queue is configured, the current scene and settings can be sent to the cluster manager for batch processing. Once a job is created, the current VREDPro instance is no longer part of the rendering process and can be closed or used for other work. Jobs in a cluster queue are processed in order of creation.
A cluster queue is a special configuration. A queue host must be defined. A VREDPro instance is started to process render jobs on the computer in use. A cluster configuration can be loaded to increase the render speed. Nodes involved in the render process must be booked by an administrator.
Using a local host allows you to do background rendering. If the queue host is localhost and you are logged into a cluster manager on localhost, rendering is done on the same PC with no additional VREDPro licenses required. For local rendering queues, no booking is required.
Cluster render jobs can be created from the render settings, the render queue or from the Cluster Manager.
When a render job that you own has finished, you can download the result images by clicking the Disk symbol. After downloading, the job is removed. Job projects and the resulting images are encrypted and stored on the cluster manager PC and only accessible by the owner. An administrator can remove jobs, but can't download the results.
Deleting a job removes all results. Deleting a job that is running cancels the process. An administrator can delete all jobs; however, you can only delete jobs that you own.
For problems with the cluster manager, see Cluster Manager Troubleshooting.