Jupyter-Cloud: state of the service

The reliability of the service Jupyter-Cloud (https://jupyter-cloud.gwdg.de) is degraded after having suffered several outages and limitations in the final weeks of 2025. At present the service operates normally but the underlying issue remains unresolved.

The symptoms

At undetermined times and under unknown circumstances users newly trying to log into the service where unable to start their notebook server (timeout) and thus unable to access their files or conduct any kind of work. Affected users had no means of influencing the situation.

Countermeasures

When a disruption became apparent affected parts of the service and sometimes the service as a whole was restarted, usually resolving the blockage. Often the reaction followed quickly, sometimes less so.

Additionally, in order to provide users with an alternative to Jupyter-Cloud as soon as possible in case a solution can not be found quickly the introduction of CoCalc (with Jupyter Notebooks as one of its components) was fast-tracked and launched in December.

Alongside the launch of CoCalc a guide was published to help users migrate data to CoCalc and recently a backup node of Jupyter-Cloud was created to allow access to user files in a minimal environment even when the main service should be impaired, for example to facilitate data transfer to CoCalc.

Outlook

GWDG Jupyter-Cloud has seen quite some groth over 2025 and the service’s resources have been scaled accordingly. Resource consumption is continuously monitored and by conventional metrics the service has enough spare capacity so service all users during peak times. Resource exhaustion is not a likely cause of recent disruptions.

The phases of unreliability have caused disruptions for individual users, classes and projects, which is very disappointing.

With the introduction of CoCalc and an Jupyter backup node to allow access to user data in any circumstances users have alternatives to access and work with their data should Jupyter-Cloud not meet demands anymore.

Regarding the Jupyter-Cloud service itself, the following steps are under consideration or planning:

1. An overhaul of the entire service. This will require a several hours long downtime.
2. A shift of Jupyter-Cloud to Kubernetes infrastructure. This requires more preparation is not suitable as a short-term solution.
3. A general recommendation to users to shift their work to CoCalc.
4. Discontinuation of Jupyter-Cloud in favor of CoCalc.

With these options and steps in mind, the situation is as follows:

• With the general availability of CoCalc shifting work to the CoCalc service is a good course of action.
• An overhaul of Jupyter-Cloud in the near future is likely.
• Any downtimes due to maintenance or overhauls will be announced accordingly.
• Should a decision be made regarding the future of the Jupyter-Cloud service an announcement will be made well ahead of time for users to continue their work in some alternative way.