Fix failing https.txt test

[autophagy]

This error occurs in some python version tests but not in others - I believe the actual exception message differs across different library versions, but the outcome is ultimately the same.

[amotl]

After just integrating #387, I've (accidentally/short-sightedly) pressed the "Update branch" button here. I don't know if that was the right decision as it created a merge commit here.

Sorry so much if that will create any hassle for you. We might want to rebase onto master and force push here again if that might resolve something.

[chaudum]

Is there anything else in the traceback that would indicate that the certificate is invalid? Asserting only the ConnectionError seems a bit generic.

[amotl]

On this matter, I just received this error with traceback (from [1]):

File "D:\a\crate-python\crate-python\src\crate\client\doctests\https.txt", line 80, in https.txt
Failed example:
    client.server_infos(crate_host)
Expected:
    Traceback (most recent call last):
    ...
    crate.client.exceptions.ConnectionError: Server not available, exception: ...[SSL: ...
Got:
    Traceback (most recent call last):
      File "C:\hostedtoolcache\windows\Python\3.7.9\x64\lib\doctest.py", line 1337, in __run
        compileflags, 1), test.globs)
      File "<doctest https.txt[11]>", line 1, in <module>
        client.server_infos(crate_host)
      File "D:\a\crate-python\crate-python\src\crate\client\http.py", line 402, in server_infos
        response = self._request('GET', '/', server=server)
      File "D:\a\crate-python\crate-python\src\crate\client\http.py", line 502, in _request
        "Server not available, exception: %s" % ex_message
    crate.client.exceptions.ConnectionError: Server not available, exception: HTTPSConnectionPool(host='localhost', port=65534): Max retries exceeded with url: / (Caused by ProtocolError('Connection aborted.', ConnectionResetError(10054, 'An existing connection was forcibly closed by the remote host', None, 10054, None)))

The particular exception is

ConnectionResetError(10054, 'An existing connection was forcibly closed by the remote host', None, 10054, None)

so there is nothing in there which might indicate the certificate is invalid, unless this error got masked by any other phenomenon.

[1] https://github.com/crate/crate-python/pull/388/checks?check_run_id=1348073153#step:5:156

[amotl]

unless this error got masked by any other phenomenon

Please forget about the gist of my last comment. The reason for the ConnectionResetError is coming from the fact that CrateDB has not spun up completely, see also #388 (comment).

However, maybe that is also the reason for the flakyness within other test suite invocations? If so, increasing wait time from 30 seconds to 45 seconds might also help, as I've found out during my work on #388.

crate-python/src/crate/testing/layer.py

Line 348 in 912acda

if wait_time > 30:

[autophagy]

The problem is, this failing https test doesnt actually set up a cratedb cluster to do it. It just creates a generic https server and attempts to query it with the client, see: https://github.com/crate/crate-python/blob/master/src/crate/client/tests.py#L210

[amotl]

I see. Thanks!

[amotl]

While I haven't been able to spot this error when running the test suite on my machine, I received respective hiccups on CI and reported about them at #386 (comment).

@autophagy: Can you look at the comment by @chaudum and then let us aim at getting this merged?

[autophagy]

@chaudum @amotl Sorry for not looking into this a little deeper - i've pushed a commit that should fix the test without modifying the expected error message.

[amotl]

I've pushed a commit that should fix the test without modifying the expected error message.

Ha. Was it really just about adding verify_ssl_cert=True?

It just creates a generic https server, see: https://github.com/crate/crate-python/blob/master/src/crate/client/tests.py#L210

I would have expected to improve on the flakeyness then by either giving it some more time to spin up at

crate-python/src/crate/client/tests.py

Line 251 in 1f3189e

time.sleep(1)

or otherwise wait for the socket being actually responsive like [1] (just to get the idea, here it is HTTP-based) or similar to [2,3] (maybe better because it is TCP-based already)?

[1]

crate-python/src/crate/testing/layer.py

Lines 360 to 371 in 1f3189e

	def _wait_for_start(self):
	"""Wait for instance to be started"""
	# since crate 0.10.0 http may be ready before the cluster
	# is fully available block here so that early requests
	# after the layer starts don't result in 503
	def validator():
	try:
	resp = self.conn_pool.request('HEAD', self.http_url)
	return resp.status == 200
	except Exception:
	return False

[2] https://github.com/lovelysystems/lovely.testlayers/blob/0.7.0/src/lovely/testlayers/util.py#L6-L13
[3] https://github.com/lovelysystems/lovely.testlayers/blob/0.7.0/src/lovely/testlayers/server.py#L77-L88

[amotl]

We can merge this anyway and see if the issue will come back again. Thanks!