When the ray cluster runs for a long time or execute many tasks, log files can be large, such as monitor.err, full of could not find ip for client xxxx. I find parameters logging_level, logging_format and log_to_driver in ray.init(). Is there a way to restrict the size of log files by default, like using logging.handlers.RotatingFileHandler and set maxBytes?

![image](https://user

How you are using LightGBM?

LightGBM component: R package

Environment info

Operating System: macOS 10.14

C++ compiler version: gcc 8.1.0

CMake version: 3.17.3

R version: 4.0.2

LightGBM version or commit hash: https://github.com/microsoft/LightGBM/tree/c07644d1d71540204a9b56f26667e8180bd009e2

Reproducible example(s)

Thanks to @Laurae2 for sharing this with m

.join seems to not work properly with unicode arrays. Here is a reproducer:

In [16]: str_arr = np.array(["hi", "there"])

In [17]: ",".join(str_arr)
Out[17]: 'hi,there'
In [18]: numba.njit(lambda str_arr: ",".join(str_arr))(str_arr)
TypingError: Failed in nopython mode pipeline (step: nopython frontend)
- Resolution failure for literal arguments:
No implementation of function 

See optuna/optuna#1913.

Some type hints are not complying to the latest mypy == 0.790 and should be addressed.

Version of Singularity:

3.0.3

Expected behavior

Some progress or update information during the creation of the SIF file.

Actual behavior

During "Creating SIF file..." step, the console does not produce any visible output for the user - but does create the resulting file eventually.

Steps to reproduce behavior

I was executing a few “larger” builds through the bui

Explain what you would like to see improved

We had CI builds stuck in downloading dependencies, e.g. VecCore. That should timeout.

Optional: share how it could be improved

Add the TIMEOUT parameter to all ExternalProject calls downloading sources.

Feature request

The asynchronous functionality of parallel-ssh is very much appreciated for management tasks. Thank you for this!

For the feature request:

It seems that the open_session() can keep worker allocated for much longer than expected in failure cases. Would it be reasonable to stretch the timeout option of wait_socket() to open_session() ?