Error saving changes on Job Cluster

[rbarrero]

Hello all and thanks.

After apply to serving a model, I go to edit corresponding Job Cluster to configure its init_script but when I try to save changes (Confirm and restart) it thrown the following error:

Error: Cannot edit cluster 0503-141315-hu3wd4ix created by ModelServing

Best regards.

[shan_chandra]

Just thought of checking,

1. Are you trying to add init script to a model serving cluster?
2. Is this issue happening on a community edition or a specific workspace ?

[rbarrero]

Hello Shan and thanks.

Yes, Im traying to add the init_script and saving changes to the cluster.

It's on a specific workspace.

[Prabakar]

If you are not able to edit the cluster to add the init script, could you please try with the global init script? This doesn't need the cluster to be edited.

[rbarrero]

Hi. I had no time to try with global init script approach. I will tell you when I try it.

Thank you for your help. Best regards.

[shan_chandra · Accepted solution]

@roberto rodriguez​  - could you please raise a support ticket with us to enable init scripts on model serving clusters on the specific workspace?

[rbarrero]

Sure, thank you.

[shan_chandra]

@roberto rodriguez​  - Just thought of following up on this, could you please let us know if you had a chance to raise a support ticket with us to enable init scripts on model serving clusters on the specific workspace? Also, please let us know if you are using a Databricks community edition or a paid service?

[User16764241763]

Hello @roberto rodriguez​ 

Could you please create the INIT script in the following location? Model Serving clusters will pick up the script in this location at start up.

dbfs:/databricks/scripts/databricks-model-serving-shared-init.sh

We may also need to enable your workspace for INIT script serving. Please file a support request with us,

[rbarrero]

Sorry for the delay in responding. Finally a partner could fix the problem, he can edit without problems the cluster and add the init_script.

Thank you!