Introduction to Monitoring¶
UbiOps has the following features to monitor your models.
Logging¶
All the logs that are created by deployments, pipelines and training runs can be found on the Logging page. You can find specific logs for deployments and pipelines by clicking their LOGS button, or by using the filter system on the Logging page.
Metrics & request overview¶
Metrics and Requests can be found under Monitoring in the left navigation panel. This will show an overview of project level metrics and requests. It is possible to filter metrics on date, or on properties like deployment_name. You can also create custom metrics for your deployment, pipeline, or training run.
Webhooks¶
Webhooks make it possible to trigger follow-up actions in the rest of your stack. You could for example sent out notifications to your team if a deployment and/or pipeline request has failed. You can configure webooks using the UbiOps WebApp (under Monitoring in the left navigation panel), Python Client Library and API.
Audit events¶
UbiOps provides four ways in which you can check your activities. The Audit events shows your activity on project level. Here you can get information about the deployments and pipelines in your currently selected project. For deployments, you can get information about when you created a deployment, updated it by uploading a new revision for a specific version or deleted a deployment (version). You get the same information for pipelines.
Note: The creation of direct and batch requests are not stored as an Audit event, as are file up- and downloads to/from your bucket(s).
Notifications¶
You can enable Notifications by configuring one or more email addresses on a deployment or pipeline version. The notification will inform you about the requests that are made to your deployment and pipelines.