diff --git a/pyspark-notebook/Dockerfile b/pyspark-notebook/Dockerfile
index e9578b19..00bb8f01 100644
--- a/pyspark-notebook/Dockerfile
+++ b/pyspark-notebook/Dockerfile
@@ -53,6 +53,10 @@ RUN cp -p "${SPARK_HOME}/conf/spark-defaults.conf.template" "${SPARK_HOME}/conf/
     echo 'spark.driver.extraJavaOptions -Dio.netty.tryReflectionSetAccessible=true' >> "${SPARK_HOME}/conf/spark-defaults.conf" && \
     echo 'spark.executor.extraJavaOptions -Dio.netty.tryReflectionSetAccessible=true' >> "${SPARK_HOME}/conf/spark-defaults.conf"
 
+# Configure IPython system-wide
+COPY ipython_kernel_config.py "/etc/ipython/"
+RUN fix-permissions "/etc/ipython/"
+
 USER ${NB_UID}
 
 # Install pyarrow
diff --git a/pyspark-notebook/ipython_kernel_config.py b/pyspark-notebook/ipython_kernel_config.py
new file mode 100644
index 00000000..8aac8fc7
--- /dev/null
+++ b/pyspark-notebook/ipython_kernel_config.py
@@ -0,0 +1,13 @@
+# Configuration file for ipython-kernel.
+# See <https://ipython.readthedocs.io/en/stable/config/options/kernel.html>
+
+# With IPython >= 6.0.0, all outputs to stdout/stderr are captured.
+# It is the case for subprocesses and output of compiled libraries like Spark.
+# Those logs now both head to notebook logs and in notebooks outputs.
+# Logs are particularly verbose with Spark, this why we turn them off through this flag.
+# <https://github.com/jupyter/docker-stacks/issues/1423>
+
+# Attempt to capture and forward low-level output, e.g. produced by Extension
+#  libraries.
+#  Default: True
+c.IPKernelApp.capture_fd_output = False  # noqa: F821