For Domino apps and workspaces you can increase the payload size by increasing the setting cerebro.domino.computegrid.kubernetes.apps.nginx.clientBodyMaxSizeMiB within Admin's Central Config.
However, this does not apply to models. If it is critical to increase this settings for a model we recommend changing the payload for the model when it is running. This means when you publish a new version the same change will need to be reapplied.
To edit the nginx configuration for specific running model use the below steps:
- Login to central and get root access using sudo su
- Get the version id for your model version by navigating to the versions tab of your model API.
- Find the right configmap of your model API
- kubectl get configmap --all-namespaces | grep <model-version-id>
- To view the configuration settings -
- kubectl describe configmap <config-map-name> -n <namespace>
- In the output, you should be able to see a key "client_max_body_size" which gives your current limit. This value can be edited using an editor
- KUBE_EDITOR="nano" kubectl edit configmap <config-map-name> -n <namespace>
- Any changes to this setting will only be available for the current running model and on publishing a new version will override this setting to the default value (10 mb). You will have to follow the above steps to edit the configmap again.
- If the request sizes are large, it can affect the logging performance. We recommend not switching on instance logging for these models
- Please remember that changing the limits to large values and hitting the model with high frequency can affect the performance of the system. Please exercise due caution and increase the limit as conservatively as possible and please do so in small increments