[
https://issues.apache.org/jira/browse/GEODE-5293?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16503970#comment-16503970
]
Ryan McMahon commented on GEODE-5293:
-------------------------------------
After a closer look, we see the following output in the stderr of the test:
```
[1528316677.446][SEVERE]: CreatePlatformSocket() returned an error, errno=97:
Address family not supported by protocol (97) Jun 06, 2018 8:24:37 PM
org.openqa.selenium.remote.ProtocolHandshake createSession INFO: Detected
dialect: OSS [1528316977.581][SEVERE]: Timed out receiving message from
renderer: 300.000 [1528316977.587][SEVERE]: Timed out receiving message from
renderer: -0.006
```
A few known chrome driver bugs appear to point to the same issue, although it
isn't entirely clear if the issue is fixed in the latest version. One thread
post indicated upgrading to 2.37 resolved the issue, however, and we are
currently on 2.35. In addition, we are pulling in the latest
google-chrome-stable which is `67.0.3396.79-1` and our chrome driver version
2.35 only supports v62-64, so that is an issue in and of itself. With this new
information, I suggest that we upgrade to the latest chrome driver and see if
that resolves the issue.
Old ticket: https://bugs.chromium.org/p/chromedriver/issues/detail?id=402
New ticket: https://bugs.chromium.org/p/chromedriver/issues/detail?id=2241
> Improve logging (and possibly robustness) of Pulse UITests
> ----------------------------------------------------------
>
> Key: GEODE-5293
> URL: https://issues.apache.org/jira/browse/GEODE-5293
> Project: Geode
> Issue Type: Test
> Reporter: Patrick Rhomberg
> Priority: Major
>
> The UITests fail intermittently in the {{WebDriverRule}}, timing out during a
> simple GET of the pulse login page. The root cause of these failures is
> difficult to identify, and the failure in general difficult to intentionally
> reproduce. Test logging should be improved to help narrow in on the actual
> cause.
> Some of the Jetty / Selenium / Chrome connections are performed in the
> constructor of the related {{ServerRule}} class, rather than in a {{before}}
> method. While the latter may be more robust in testing, the former may
> actually reveal deficiencies in Pulse server production code.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
