Uploaded image for project: 'JBoss Enterprise Application Platform'
  1. JBoss Enterprise Application Platform
  2. JBEAP-1767

IPv6 UDP multicast: JGroups Failed to start service and Undertow mod_cluster proxy problems

    XMLWordPrintable

Details

    Description

      Dear rhn-engineering-bban and sdouglas1@redhat.com, I encountered problems with UDP multicast on IPv6, intimately related to my earlier report on UNDERTOW-454.

      Facts

      • all servers bound to global scope IPv6 address 2620:52:0:105f::ffff:0
      • test runs on a single box, on a one NIC
      • pure IPv6 RHEL6 x86_64, i.e. no IPv4 interfaces configured (not even localhost)

      Configurations

      UDP multicast config matrix

      • GLOBAL = global scope UDP multicast, starts with ff0e::*
      • NETWORK_LOCAL = local network nodes UDP multicast, starts with ff02::*
      • INTERFACE_LOCAL = interface local UDP multicast, starts with ff01::*

      EAP 6.4.0 and EAP 7.0.0.DR13 config

      • EAP 7.0.0.DR13 acts both as worker and balancer, depending on the configuration; see Section Results below
      Interfaces

      The same for all the test runs:

      <interface name="management">
      <inet-address value="2620:52:0:105f::ffff:0"/>
      </interface>
      <interface name="public">
      <inet-address value="2620:52:0:105f::ffff:0"/>
      </interface>
      
      UDP multicast config matrix
      • GLOBAL
        <socket-binding name="jgroups-mping" port="0" multicast-address="ff0e::12" multicast-port="45700"/>
        <socket-binding name="jgroups-udp" port="55200" multicast-address="ff0e::12" multicast-port="45688"/>
        <socket-binding name="modcluster" port="0" multicast-address="ff0e::a" multicast-port="35432"/>
        
      • NETWORK_LOCAL
        <socket-binding name="jgroups-mping" port="0" multicast-address="ff02::0" multicast-port="45700"/>
        <socket-binding name="jgroups-udp" port="55200" multicast-address="ff02::0" multicast-port="45688"/>
        <socket-binding name="modcluster" port="0" multicast-address="ff02::a" multicast-port="35432"/>
        
      • INTERFACE_LOCAL
        <socket-binding name="jgroups-mping" port="0" multicast-address="ff01::14" multicast-port="45700"/>
        <socket-binding name="jgroups-udp" port="55200" multicast-address="ff01::14" multicast-port="45688"/>
        <socket-binding name="modcluster" port="0" multicast-address="ff01::a" multicast-port="35432"/>
        

      Apache HTTP Server config

      • either from JWS 3.0.2-ER1 or EWS 2.1.0
      • Listen [2620:52:0:105f::ffff:0]:8847
      • GLOBAL AdvertiseGroup [ff0e::a]:35432
      • NETWORK_LOCAL AdvertiseGroup [ff02::a]:35432
      • INTERFACE_LOCAL AdvertiseGroup [ff01::a]:35432

      Results

      UDP Multicast Workers Balancer Result Comment
      GLOBAL EAP 6.4.0 EWS 2.1.0 No comment.
      NETWORK_LOCAL EAP 6.4.0 EWS 2.1.0 harmless SocketException
      INTERFACE_LOCAL EAP 6.4.0 EWS 2.1.0 No comment.
      GLOBAL EAP 7.0.0.DR13 JWS 3.0.2-ER1 No comment.
      NETWORK_LOCAL EAP 7.0.0.DR13 JWS 3.0.2-ER1 JGroups don't start MSC000001
      INTERFACE_LOCAL EAP 7.0.0.DR13 JWS 3.0.2-ER1 No comment.
      GLOBAL EAP 7.0.0.DR13 Undertow 1.3.3.Final No exceptions; balancer advertising doesn't work
      NETWORK_LOCAL EAP 7.0.0.DR13 Undertow 1.3.3.Final JGroups don't start MSC000001
      INTERFACE_LOCAL EAP 7.0.0.DR13 Undertow 1.3.3.Final No exceptions; balancer advertising doesn't work
      Regarding "advertising doesn't work"

      The EAP 7.0.0.DR13 Undertow 1.3.3.Final balancer seems to be working, according to the log:

      INFO  [io.undertow] (MSC service thread 1-1) UT005039: Undertow starts mod_cluster proxy advertisements on /ff0e:0:0:0:0:0:0:a:35432 with frequency 10000 ms
      ...snip...
      09:37:53,018 DEBUG [io.undertow] (default I/O-1) UT005040: Gonna send payload:
      HTTP/1.0 200 OK
      Date: Thu, 5 Nov 2015 09:37:53 -0500
      Sequence: 1
      Digest: 70074bf470f252362639a18364e27a45
      Server: 2b0c7b3b-3869-41f4-bcfc-923da9f774a9
      X-Manager-Address: 2620:52:0:105f:0:0:ffff:0%eth0:8484
      X-Manager-Url: /
      X-Manager-Protocol: http
      X-Manager-Host: 2620:52:0:105f:0:0:ffff:0%eth0

      Workers also seem to be in order, up and running:

      INFO  [org.jboss.modcluster] (ServerService Thread Pool -- 65) MODCLUSTER000032: Listening to proxy advertisements on /ff0e:0:0:0:0:0:0:a:35432

      None of the workers ever replies to the balacner though. Any suggestions, sdouglas1@redhat.com?

      Please, share your view on the topic. Thanks guys.

      Attachments

        Issue Links

          Activity

            People

              rhn-engineering-rhusar Radoslav Husar
              mbabacek1@redhat.com Michal Karm
              Stuart Douglas
              Michal Karm Michal Karm
              Michal Karm Michal Karm
              Votes:
              0 Vote for this issue
              Watchers:
              9 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: