Github user neykov commented on a diff in the pull request:
https://github.com/apache/brooklyn-server/pull/560#discussion_r101520554
--- Diff:
launcher/src/main/java/org/apache/brooklyn/launcher/BrooklynWebServer.java ---
@@ -447,7 +450,15 @@ public synchronized void start() throws Exception {
rootContext.setTempDirectory(Os.mkdirs(new File(webappTempDir,
"war-root")));
server.setHandler(handlers);
- server.start();
+ try {
+ server.start();
+ } catch (BindException e) {
+ // port discovery routines may take some time to clear, e.g.
250ms for SO_TIMEOUT
+ // tests fail because of this; see if adding a delay improves
things
+ log.warn("Initial server start-up failed binding (retrying
after a delay): "+e);
+ Time.sleep(Duration.millis(500));
+ server.start();
--- End diff --
> Why would setReuseAddress(true) improve false positives? As long as we
make the same call when we start the server it should be identical? (Or are you
thinking that the root cause is that the server-start call isn't using that,
and that is what you'd fix?)
That's right - `setReuseAddress(true)` is used in `isPortAvailable`, but
not in the actual server bind call (for web or jmx). `SO_REUSEADDR` allows you
to bind to a port in `TIME_WAIT` state. So the availability check will succeed,
but the server bind will fail.
I wouldn't add `setReuseAddress(true)` to the server, but will instead
remove it from `isPortAvailable`.
We have parallel tests not from the same maven run, but from parallel
jenkins jobs (another process).
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
