To find out if this integration is available in your organization, see your Datadog Integrations page or ask your organization administrator.
To initiate an exception request to enable this integration for your organization, email support@ddog-gov.com.
Run your generative AI models, large-scale batch jobs, job queues, and much more through Modal.
See audit logs, function logs, and utilization metrics in the Modal Datadog dashboard
Overview
Modal lets you run generative AI models, large-scale batch jobs, job
queues, and much more. This integration collects logs (stdout/stderr logging
from your modal applications, or audit logs for your account) that you can
visualize through Datadog dashboards and set up alerts for with Datadog
monitors.
Setup
Installation
To set up the Modal integration:
Navigate to the Modal tile on the Datadog Integrations page and
click Install Integration.
Click Connect Accounts to begin authorization of this integration. You
will be redirected to log into Modal, and once logged in, you’ll be
redirected to the Datadog authorization page.
Click the Authorize button to complete setup.
Logs and metrics from your Modal apps should now start appearing in Datadog.
Data Collected
Logs
Modal collects all application (stdout/stderr logging for functions) and audit logs.
Events
Modal does not include any events.
Metrics
| |
|---|
modal.container.running
(gauge) | Number of running containers
Shown as container |
modal.container.terminations
(count) | Number of terminated containers |
modal.cpu.requested
(gauge) | Number of CPU cores requested |
modal.cpu.utilization
(gauge) | Number of CPU cores in use |
modal.gpu.compute.utilization
(gauge) | GPU compute utilization |
modal.gpu.count
(gauge) | Number of GPUs attached to the container (zero for CPU containers) |
modal.gpu.memory.usage
(gauge) | Amount of GPU memory in use
Shown as byte |
modal.gpu.memory.utilization
(gauge) | GPU memory utilization
Shown as mebibyte |
modal.gpu.power.usage
(gauge) | GPU power usage
Shown as watt |
modal.gpu.power.utilization
(gauge) | GPU power utilization as a fraction of TDP |
modal.gpu.temperature
(gauge) | GPU temperature
Shown as degree celsius |
modal.input_events.elapsed_time_us
(count) | Elapsed time of input events
Shown as microsecond |
modal.input_events.successes
(count) | Number of successful completed inputs |
modal.input_events.total_inputs
(count) | Number of completed inputs |
modal.memory.requested
(gauge) | Amount of memory requested
Shown as byte |
modal.memory.usage
(gauge) | Amount of memory in use
Shown as byte |
modal.memory.utilization
(gauge) | Memory in use
Shown as mebibyte |
modal.gpu.count reports the number of GPUs attached to each container, or zero
for CPU containers. Container resource metrics from GPU containers include the
gpu_billing_type tag when the billing type is recognized. This tag identifies the
billed GPU class (for example, H100), which may differ from the physical GPU model
when a workload is scheduled onto a newer GPU.
Uninstallation
Once this integration has been uninstalled, any previous authorizations are
revoked and logs/metrics stop being emitted to Datadog.
- Navigate to the Modal metrics settings page
and select Delete Datadog Integration.
- On the Configure tab in the Modal integration tile in Datadog,
click Uninstall Integration.
- Confirm that you want to uninstall the integration.
- Ensure that all API keys associated with this integration have been
disabled by searching for the integration name on the API Keys
page.
Troubleshooting
Need help? Contact Modal support.
