In the Promote UI, select your username in the top right corner, and then select Admin from the dropdown menu. There, you can find these sub-tabs that allow you to manage your Promote instance:
- System Overview
The Models tab provides you with a list of all models associated with your Promote instance.
Every column is searchable.
The Users tab shows a list of all users associated with your Promote instance. This includes information like the user's email and API key. For more information about what you can do through the Users tab, see Manage Users.
The System Overview tab provides you with an overview of your Promote system. It contains this information:
- CPU usage
- RAM usage
- Internet connectivity status
- Disk-usage information
If you see a green check mark next to those pieces of information, it means the system is operating as expected.
The Advanced tab contains options for these:
We strongly recommend that you leave the base images for Python and R as DEFAULT. Alteryx must approve the change and upload the base image in Quay beforehand.
You can save the Base Images for your Python or R models: paste the base-image reference into the appropriate field, then select Save. You must redeploy the model for the change to take effect.
You can set the Disk Bundle Limit. This restricts the number of bundles that you can save on disk per model, because bundles can take up a significant amount of space.
To set the limit, specify the number of bundles you want to limit Promote to. Select Save.
Setting your bundle limit may affect your ability to revert to a previous version.
You can decide whether Promote should save detailed logs about predictions for models in Development / Staging and Production. Toggle the switches to decide if Promote should create logs for those. When turned on, you can view the logs in the History section.
By default, Promote stores logs of every prediction request for up to 14 days, unless the logs exceed 50 GB of storage. In that case, the system indexes the logs by day, then prunes older days until it reaches safe storage levels.
Any time you toggle on or off the log switches, allow a few minutes for the system to update. You may experience a few seconds of prediction downtime as the system updates each model.