Tests which fail in multiprocessing contexts

nbconvert/preprocessors/tests/files/Sleep One.ipynb

@@ -0,0 +1,50 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import time"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "time.sleep(0.01)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.7.0"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

nbconvert/preprocessors/tests/test_execute.py

@@ -309,6 +309,74 @@ def test_parallel_notebooks(capfd, tmpdir):
     captured = capfd.readouterr()
     assert captured.err == ""
 
+@pytest.mark.xfail
+def test_many_parallel_notebooks(capfd):
+    """Ensure that when many IPython kernels are run in parallel, nothing awful happens.
+
+    Specifically, many IPython kernels when run simultaneously would enocunter errors
+    due to using the same SQLite history database.
+    """
+
+    # I've put timeout=5, which is a bit aggressive, but if I destroy the ZMQ context
+    # below, timeout=5 works well enough.
+    opts = dict(kernel_name="python", timeout=5)
+    input_name = "HelloWorld.ipynb"
+    input_file = os.path.join(current_dir, "files", input_name)
+    res = PreprocessorTestsBase().build_resources()
+    res["metadata"]["path"] = os.path.join(current_dir, "files")
+
+    # run once, to trigger creating the original context
+    run_notebook(input_file, opts, res)
+
+    # Destroy the context - if you don't do this, the context
+    # will survive across the fork, and then fail to start properly.
+    import zmq
+    zmq.Context.instance().destroy()
+
+    with mp.Pool(4) as pool:
+        pool.starmap(run_notebook, [(input_file, opts, res) for _ in range(8)])
+
+    captured = capfd.readouterr()
+    assert captured.err == ""
+
+@pytest.mark.xfail
+def test_parallel_fork_notebooks(capfd):
+    """Ensure that when many IPython kernels are run in parallel, nothing awful happens.
+
+    Specifically, many IPython kernels when run simultaneously would enocunter errors
+    due to using the same SQLite history database.
+    """
+
+    opts = dict(kernel_name="python", timeout=5)
+    input_name = "Sleep One.ipynb"
+    input_file = os.path.join(current_dir, "files", input_name)
+
+    fast_name = "HelloWorld.ipynb"
+    fast_file = os.path.join(current_dir, "files", fast_name)
+
+
+    res = PreprocessorTestsBase().build_resources()
+    res["metadata"]["path"] = os.path.join(current_dir, "files")
+
+    # run once, to trigger creating the original context
+    thread = threading.Thread(target=run_notebook, args=(input_file, opts, res))
+    thread.start()
+
+    try:
+        # Destroy the context - if you don't do this, the context
+        # will survive across the fork, and then fail to start properly.
+        # but if you do do this, then the context will get destroyed
+        # while the kernel is running in the current thread.
+        import zmq
+        zmq.Context.instance().destroy()
+
+        with mp.Pool(4) as pool:
+            pool.starmap(run_notebook, [(fast_file, opts, res) for _ in range(8)])
+    finally:
+        thread.join(timeout=1)
+
+    captured = capfd.readouterr()
+    assert captured.err == ""
 
 class TestExecute(PreprocessorTestsBase):
     """Contains test functions for execute.py"""