View worker details in the worker dashboard
The worker dashboard provides details for the worker that processes a task. You can see:
-
Metadata, such as the instance type, for the worker
-
The session actions that the worker performed
-
Worker performance, including CPU, memory and disk usage
-
A graph of the CPU, memory, and disk usage over time
-
A graph of the disk speed over time
-
The worker log for the task
To view the worker dashboard from a task
-
Follow the steps in View and manage job details in Deadline Cloud to view a list of jobs.
-
Select a job from the Jobs list.
-
Select a step from the Steps list.
-
Select a task from the Tasks list.
-
In the task table, from the Actions menu, choose View worker dashboard.
To view the worker dashboard from fleet details
-
Follow the steps in View queue and fleet details in Deadline Cloud to view a fleet.
-
Select a Worker from the Workers list.
-
From the Actions menu, choose View worker dashboard.
Use cases
Detecting under-provisioned instances
When renders take longer than expected, the worker dashboard can help determine if your instances are adequately sized for your workloads. While 100% vCPU utilization is normal for many renderers, consistently high memory usage near maximum capacity and elevated disk space utilization may indicate that your instances are under-provisioned. In such cases, upgrading your fleet's instance configuration can reduce render errors and significantly improve render times. However, it's important to continue monitoring worker performance after upgrading to ensure you've found the optimal balance - upgrading too aggressively can lead to unnecessary costs through over-provisioning.
Detecting over-provisioned instances
Even when tasks are completing successfully, there may be opportunities to optimize your costs. The worker dashboard can reveal if you're paying for more compute power than your workloads require. If you see that the worker has low average vCPU usage, minimal memory utilization, and excess unused disk space, you can downsize the instance configuration of your fleet.
Troubleshooting failed tasks
When investigating failed tasks, the worker dashboard serves as a valuable diagnostic tool. Pay particular attention to peak memory usage and disk space utilization - if these metrics approach or reach 100%, they're likely the root cause of your task failures. Such resource exhaustion indicates that your current instances lack the capacity to handle your workloads effectively. In these cases, provisioning instances with increased memory or disk space will help ensure successful task completion.
Optimal instance utilization rate
vCPU Utilization
Target range: 70–90%
-
Below 70%: Likely underutilizing compute resources, meaning you're paying for more CPU than your workload needs
-
70–90%: Optimal range where you're efficiently using resources without hitting bottlenecks
-
Consistently at 100%: Could indicate CPU bottlenecks that might slow down renders
Keep in mind that some render tasks will naturally be more CPU-intensive than others, and 100% vCPU usage may not be an issue. Real-time visualization tasks might show more consistent CPU utilization, while tasks with changing computational requirements might have varying patterns.
Memory Utilization
Target range: 70–85%
-
Below 50%: Potentially oversized instances for your workload
-
70–85%: Optimal utilization with enough headroom for spikes
-
Above 90%: Risk of performance degradation or out-of-memory errors
Memory requirements can vary significantly depending on scene complexity, texture resolution, and simulation data. Monitoring memory trends over time is important to identify if your workloads are growing in memory requirements.
Disk Space Utilization
Target range: 60–80%
-
Below 40%: Likely over-provisioned storage
-
60–85%: Good utilization with room for temporary files and caches
-
Above 85%: Risk of running out of space during large renders
Remember that disk I/O performance can be just as important as capacity, especially for workloads that read/write large texture or cache files during rendering.
