[
https://issues.apache.org/jira/browse/GEODE-8241?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17133416#comment-17133416
]
ASF GitHub Bot commented on GEODE-8241:
---------------------------------------
aaronlindsey commented on a change in pull request #5236:
URL: https://github.com/apache/geode/pull/5236#discussion_r438937359
##########
File path:
geode-membership/src/integrationTest/java/org/apache/geode/distributed/internal/membership/gms/MembershipIntegrationTest.java
##########
@@ -172,6 +179,91 @@ public void
secondMembershipCanJoinUsingTheSecondLocatorToStart()
stop(locator2, locator1);
}
+ @Test
+ public void secondMembershipPausesForLocatorWaitTime()
+ throws IOException, MemberStartupException {
+
+ /*
+ * Start a locator for the coordinator (membership) so we have a port for
it.
+ *
+ * Its locator-wait-time is set to 0 so it eventually (soon after
membership is started) forms a
+ * distributed system and becomes a coordinator.
+ */
+
+ final MembershipLocator<MemberIdentifier> coordinatorLocator =
createLocator(0);
+ coordinatorLocator.start();
+ final int coordinatorLocatorPort = coordinatorLocator.getPort();
+
+ final Membership<MemberIdentifier> coordinatorMembership =
+ createMembership(coordinatorLocator, coordinatorLocatorPort);
+
+ /*
+ * We have not even started the membership yet — connection attempts will
certainly fail until
+ * we do. This is a bit like the locator (host) not being present in DNS
(yet).
+ */
+
+ /*
+ * Start a second locator and membership trying to join via the
coordinator (membership) that
+ * hasn't yet started behind the port.
+ *
+ * Set its locator-wait-time so it'll not become a coordinator right away,
allowing time for the
+ * other member to start and become a coordinator.
+ *
+ * Calculate the locator-wait-time to be greater than the minimum wait
time for connecting to a
+ * locator.
+ */
+
+ final MembershipLocator<MemberIdentifier> lateJoiningLocator =
createLocator(0);
+ lateJoiningLocator.start();
+ final int lateJoiningLocatorPort = lateJoiningLocator.getPort();
+
+ final int[] lateJoiningMembershipLocatorPorts =
+ new int[] {coordinatorLocatorPort, lateJoiningLocatorPort};
+
+ final Duration minimumJoinWaitTime = Duration
+ .ofMillis(2_000) // expected amount of sleep time per loop in
GMSJoinLeave.join()
Review comment:
It seems like it would be better to reference the same constants that
are used in `GMSJoinLeave` instead of hard-coding this value.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> Locator does not observe locator-wait-time
> ------------------------------------------
>
> Key: GEODE-8241
> URL: https://issues.apache.org/jira/browse/GEODE-8241
> Project: Geode
> Issue Type: Bug
> Reporter: Aaron Lindsey
> Assignee: Aaron Lindsey
> Priority: Major
>
> In the case where a locator starts up and is unable to connect to any other
> locators, it may decide to become the membership coordinator even if
> locator-wait-time has not elapsed.
> The following conditional from GMSJoinLeave.java causes the issue. There
> should be an additional check for locator-wait-time before becoming
> coordinator.
> {code:java}
> if (state.joinedMembersContacted <= 0 &&
> (tries >= minimumRetriesBeforeBecomingCoordinator ||
> state.locatorsContacted >= locators.size())) {
> synchronized (viewInstallationLock) {
> becomeCoordinator();
> }
> return true;
> }
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
