Uploaded image for project: 'JGroups'
  1. JGroups
  2. JGRP-2511

JGroups Tunnel Protocol in Multi Gossip router setup disturbs cluster view

    XMLWordPrintable

Details

    • Bug
    • Resolution: Cannot Reproduce
    • Major
    • 5.1, 4.2.10
    • 4.1.4
    • None
    • False
    • False
    • Undefined

    Description

      Steps to reproduce:

      We have 2 EAP instance EAP1 and EAP2, and two gossip router, route 1 and route 2

       

      EAP 1 ---> Route1

                 ---> Route 2

       
      EAP 2 ---> Route1

                 ---> Route 2
      create 2 eap with standalone-full-ha config, change the udp jgroups replace the section for tunnel
       
      create two gossip router say port 1 and port 2
       
      start the gossip routers
       
      start the two eap
       
      u will see both eap connects to gossip router 1
       
      then stop gossip router 1
       
      u will see the eap connects to gossip router 2
       
      start gossip router 1 again
       
      now all eap still in gossip router 2
       
      stop eap1
       
      and start it again
       
      u will see eap1 connects to gossip router 1
       
      while eap2 connects to gossip router 2

      Above steps are also tested wit RHSSO 7.4 to form 2 node cluster using Tunnel Protocol with multiple gossip router.

      Cluster view is affected in both EAP/RHSSO clustering using Tunnel protocol in multiple gossip router setup.

      From http://jgroups.org/manual4/index.html#TUNNEL_Advanced

      A GossipRouter is not a single point of failure. In a setup with multiple gossip routers, the routers do not communicate among themselves, and a single point of failure is avoided by having each channel simply connect to multiple available routers. In case one or more routers go down, the cluster members are still able to exchange messages through any of the remaining available router instances, if there are any.

      In the above scenario nodes are getting connected to gossip routers randomly and as a result the cluster is affected since both nodes join different gossip router.

      As per my understanding the expected behavior of gossip router should not affect the cluster view.

      Discussed this in https://issues.redhat.com/browse/EAPSUP-290 reproduce.tar.gzgossip.xml

      Attachments

        1. gossip.xml
          0.6 kB
        2. reproduce.tar.gz
          14 kB
        3. workaround.tar.gz
          13 kB

        Activity

          People

            rhn-engineering-bban Bela Ban
            rhn-support-vasharma Varsha Sharma (Inactive)
            Votes:
            0 Vote for this issue
            Watchers:
            5 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: