UbiOps version 2.10.0
On the **20th of July 2021** we have released new functionality and made improvements to our UbiOps SaaS product. An overview of the new functionality and changes:
Support for requests up to 48 hours
Batch requests can now take up to 48 hours to complete. This enables you to run longer running workloads on UbiOps, such as training jobs, parsing, big database insertions and many more. All new and existing batch deployments support these longer running requests. For express requests, the time limit
remains 1 hour as it was before.
Canceling batch requests
It is now possible to cancel batch requests while they are in progress. This will terminate your code and release the computing resources of the deployment instance within 30 seconds. You can use the cancel button in the WebApp or use the API to cancel requests. It is for now not yet possible to cancel express requests.
Separation of batch and express request deployment types
In order to support 48-hour batch requests, we had to separate deployments into batch and express deployments. This means that from now on, either only batch or express requests can be made to a deployment. You select the deployment type when you create it. All existing deployments have been migrated to one of the two types as well. We determined which type of request was used for all the existing deployments by looking at the last 3 months of activity. We have been in touch with customers with deployments where we were unable to determine the best request type. You can find the request type of each deployment by viewing the details of the deployment in the WebApp or on the API.
Batch deployments process requests asynchronously, these batch requests are allowed to take up to 48 hours for processing. This can for instance be useful for running training jobs, optimization scripts or large database insertions.
Express deployments process requests in a synchronous way, these direct requests are useful for running tasks that generate a quick response and need to be implemented accordingly. Think of model inference, data processing steps or similar calculations.
⚠️ You can no longer send both request types (batch & direct) to the same deployment. Instead, you should choose the request type when creating a new deployment.
Depreciation of Python 3.5
We have dropped support for Python 3.5 in this release, as it has reached end-of-life status a while ago and is no longer maintained. We have been in touch with all customers still running Python 3.5 deployments and helped them migrate to a newer version.
Deployment versions limit
In the paid plans you get unlimited deployment versions. Unlimited deployment versions really means unlimited. But for practical reasons, we have set the default limit to 199.
Need more? Contact us and we’ll adjust it accordingly.