(Test sentence)

Problems with the Jupyter notebook environment ("Lab")

Error message "Error in Authenticator.pre_spawn_start: ... 401: Unauthorized"

When a new user accesses the Lab in a new collab for the first time, it is possible, that the JuypterLab is not starting and instead shows a "...401 Unauthorized" message.

To fix this error, please access the "Drive" link in the collab once (the "Drive" is also in the left hand navigation menu of the Collab). After the "Drive" has been accesses once the JupyterLab should start.

No example notebooks in the Jupyter notebook environment ("Lab")

When starting the Jupyter notebook environment for the first time it takes about a minute (without visible "in progress" indication) until the example files appear in the file-chooser. Until then the list is just empty. So the fix here is "have a minute patience"

Timeout when starting the Jupyter notebook environment ("Lab")

It is possible, that accessing the "Lab" link results in a "timeout" message. This can happen, when the lab environment is started for the first time on one of the virtual machines providing the service, as the very first start per machine takes quite some time (in the order of 2 minutes). Accessing the "Lab" link again after a few minutes usually works.

Lab error message "Cannot open - Could not find path..."

When first opening the Lab it is possible, that it can not open the path /shared/name-of-the-collab, even though in the drive the path is already accessible (see the two screenshots below from the Lab and the Drive). This is due to some delay in synchronisation between the file systems. Fix: Wait a few minutes and try again... (If that does not fix the problem, please contact support@ebrains.eu)

In the "Drive" the path is visible:

Test jobs make problems at Jülich

Since August 2021 accessing the "Lab" link in the left hand navigation menu brings up a new site selection menu:

Via choosing a different lab execution site it is be possible to choose a specific site, if there are specific needs for the created Jupyter virtual machine for a given task (e.g. a large memory requirement). Currently all the examples do not need any site specific features, so you can select in principle any site. BUT (as of end of August 21) there is a sporadic problem with the NMPI communication, if running at Jülich. Thus for now, please use CSCS as the primary execution site.

BrainScaleS specific tips

The BrainScaleS external user chat has service announcements (e.g. when a system is down due to maintenance) and is also a good place to ask general questions.

BrainScaleS-2: "Job error" in the Collaboratory, though it worked until 21 June 2021

On 21 June 2021 we changed the platform selection in the nmpi command client.submit_job(...,platform=nmpi.BRAINSCALES,...). From then on, please specify nmpi.BRAINSCALES2 as the platform to run networks on the BrainScaleS-2 architecture (i.e. on "wafer">60).

In the JobManager, a job with the now wrong platform=nmpi.BRAINSCALES shown in the "log" part of the detailed job information:

72 is not a valid wafer module for BrainScaleS-1. Since 21 June 2021, please use "nmpi.BRAINSCALES2" as platform specification ...

Also since 21 June 2021 a separate quota setting is used for BrainScaleS-2 (visible on the Quota tab in the JobManager)

BrainScaleS-1

Currently (2 July 2021) the default-system ("wafer 33") is under maintenance. Please see the BrainScaleS-1 chat for updates.