geode-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF subversion and git services (JIRA)" <>
Subject [jira] [Commented] (GEODE-3588) 2 restarts of Locator results in split brain
Date Tue, 09 Jan 2018 16:57:00 GMT


ASF subversion and git services commented on GEODE-3588:

Commit 3cf7caab3c3726dfb47e12a900240e377e035594 in geode's branch refs/heads/develop from
[;h=3cf7caa ]

Squashed commit of the following:

commit 70a76592a58379bdd9b53433877b64831fc7432e
Author: Bruce Schuchardt <>
Date:   Tue Jan 9 08:37:45 2018 -0800

    GEODE-3588 2 restarts of Locator results in split brain

    removed thread dump in new test

commit 89bf34c39f3df4ed7b16d6c9a256e2d26b9d2267
Author: Bruce Schuchardt <>
Date:   Mon Jan 8 15:57:09 2018 -0800

    GEODE-3588 2 restarts of Locator results in split brain

    Udo's fix for GEODE-870 added a new boolean instance variable to
    GMSJoinLeave to tell its ViewCreator thread to shut down.  This works
    but the state was never being reset after its first use.  This caused
    Subsequent ViewCreator threads to shut down immediately.  The only
    way to fix this condition without a patch is to restart the coordinator node.

    The patch moves this boolean variable to the ViewCreator thread so that
    it is automatically reset when a new ViewCreator is instantiated.

    I also did a little code cleanup, moving GMSJoinLeave methods from the
    end of the file to where its other methods are located and adding
    a setShutdownFlag() method during debugging so I could isolate what
    was happening.

Sarge reviewed the changes for me so this closes #1255

> 2 restarts of Locator results in split brain
> --------------------------------------------
>                 Key: GEODE-3588
>                 URL:
>             Project: Geode
>          Issue Type: Bug
>          Components: locator, membership
>            Reporter: Michael Martell
>         Attachments: RunLocalhostGemfire.ps1
> Restarting the locator and listing members works properly. Repeating the process a second
time shows no cluster members.
> To repro:
> 1) start a locator
> 2) start a server
> 3) list members (shows locator and server)
> 4) stop the locator
> 5) restart the locator (same as step 1)
> 6) list members (shows locator and server)
> 7) stop the locator a second time
> 8) restart the locator a second time (same as step 1)
> 9) list members shows only the locator
> Notes:
> - This was done on a Windows machine using latest geode/develop (as of 9/11/2017)
> - Was tried with and without using protobuf=true flag, and fails in both cases)

This message was sent by Atlassian JIRA

View raw message