Batchspawner spawning / keep-alive is instable

Hi, I often end up in the following situation:

• JupyterHub spawns new worker
• Worker starts:

+ which jupyterhub-singleuser
/opt/anaconda/3-2019.10/bin/jupyterhub-singleuser
+ batchspawner-singleuser jupyterhub-singleuser --ip=0.0.0.0 --NotebookApp.default_url=/lab
[I 2020-03-29 18:18:43.199 SingleUserNotebookApp manager:48] [nb_conda_kernels] enabled, 84 kernels found
[I 2020-03-29 18:18:44.331 SingleUserNotebookApp extension:157] JupyterLab extension loaded from /opt/anaconda/3-2019.10/lib/python3.7/site-packages/jupyterlab
[I 2020-03-29 18:18:44.331 SingleUserNotebookApp extension:158] JupyterLab application directory is /opt/anaconda/3-2019.10/share/jupyter/lab
[I 2020-03-29 18:18:44.784 SingleUserNotebookApp __init__:31] [Jupytext Server Extension] Deriving a JupytextContentsManager from LargeFileManager
[I 2020-03-29 18:18:44.788 SingleUserNotebookApp singleuser:561] Starting jupyterhub-singleuser server version 1.1.0
[I 2020-03-29 18:18:44.800 SingleUserNotebookApp notebookapp:1924] Serving notebooks from local directory: /data/nasif12/home_if12/the_user
[I 2020-03-29 18:18:44.800 SingleUserNotebookApp notebookapp:1924] The Jupyter Notebook is running at:
[I 2020-03-29 18:18:44.800 SingleUserNotebookApp notebookapp:1924] http://lab-desk12:38375/jupyter/user/the_user/
[I 2020-03-29 18:18:44.800 SingleUserNotebookApp notebookapp:1925] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).
[I 2020-03-29 18:18:44.820 SingleUserNotebookApp singleuser:542] Updating Hub with activity every 300 seconds

Also, after ~ 24h the jupyterlab-singleuser worker looses connection to the JupyterHub. However, depending on the configuration I would like to have the jupyter instances running for more than one day.

Does somebody know why the connection between JupyterHub and the worker is so instable? Could https://github.com/jupyterhub/jupyterhub/pull/2727 be the solution?

[consideRatio]

Closed by #187

[linhbngo]

@Hoeze We ran into the same issue at our cluster (OpenPBS). Since PR #187 has not been merged yet, we side-stepped by set the JupyterHub Spawner’s poll_interval configuration to 1800 seconds (default is 30 seconds), and it seemed to help.