geode-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF subversion and git services (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (GEODE-77) Replace JGroups 2.2.9
Date Thu, 24 Sep 2015 22:56:04 GMT

    [ https://issues.apache.org/jira/browse/GEODE-77?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14907195#comment-14907195
] 

ASF subversion and git services commented on GEODE-77:
------------------------------------------------------

Commit eab327f6a2b9bc2ca715464f61b4388ce304df2f in incubator-geode's branch refs/heads/feature/GEODE-77
from [~bschuchardt]
[ https://git-wip-us.apache.org/repos/asf?p=incubator-geode.git;h=eab327f ]

GEODE-77 fixes for failing unit and integration tests

  if the coordinator returned by locators doesn't work we now
  ask other members in the view returned by the locator

  leave & crash events now ack the view currently being installed
  so that it won't wait for the associated members

  views are always installed using the two-phase protocol.
  GemFire did this and some algorithms, such as rebalancing,
  are thrown off if departed members are in the view when
  a rebalance starts.

  views are always transmitted by the View Creator thread.
  GMSJoinLeave.becomeCoordinator() installs an initial view
  into the View Creator before starting it if there is a view
  to prepare and install.

  Rebalance operations targetting a new member were failing to
  achieve balance if an old ID for that member was still in the
  membership view.  We now detect those old IDs and remove them
  when the new member joins.

  ClientProxyMembershipID was not being deserialized properly
  after being transmitted from one member to another.

  HealthMonitor classes with Ping in their names are renamed to
  Check.  I've found that Ping confuses people - they
  sometimes think that gemfire is using network ping protocol

  HealthMonitor beSick/playDead were tightened up

  HealthMonitor was not recording activity from other members unless
  it was currently watching that member.  This caused a lot of
  unnecessary suspicion when switching from one member to another.

  HealthMonitor wasn't clear about who was raising suspicion and
  who was the target of the suspicion.

  FindCoordinatorRequest was using java serialization for rejected
  coordinator IDs.

  GMSMembershipManager's latestViewLock use wasn't quite the same
  as in the membership manager in GemFire 8.2 and caused a deadlock.

  fixing GEODE-360: PRTombstoneMessage should ignore
  ForceReattemptException

  LogFileParser wasn't recognizing timestamps for thread dumps, causing
  them to be mis-sorted when merging log files.


> Replace JGroups 2.2.9
> ---------------------
>
>                 Key: GEODE-77
>                 URL: https://issues.apache.org/jira/browse/GEODE-77
>             Project: Geode
>          Issue Type: Bug
>            Reporter: Bruce Schuchardt
>            Assignee: Bruce Schuchardt
>            Priority: Blocker
>             Fix For: 1.0.0-incubating
>
>         Attachments: GEODE-MembershipManagerFunctionalSpecification-130715-1604-29054.pdf
>
>
> The JGroups 2.2.9 sources that are currently included in Geode must be replaced in order
for Geode to leave incubation.  A wiki document has been created to investigate alternatives.
> https://cwiki.apache.org/confluence/display/GEODE/Replacing+JGroups+2.2.9



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message