Re: [jgroups-dev] Dynamic host problem.

This seems to be an operator error, resulting in the 2 clusters finding 
each other.

What could be done here is
- Add a (JMX) method in TCPPING to clear the dynamic_hosts cache. This 
could be invoked via jconsole, for instance.
- Add AUTH, so only authenticated members can join. This seems to be an 
overkill though...


On 4/13/11 10:41 AM, Kuwon Kang wrote:
> Hello Bela.
>
> I'm kuwon and I found some problem.
>
> We are using JGroups in production mode.
> And our team are preparing disaster recovery system for Customer system.
>
> But the guy who is in charge of the disaster recovery system that he copied
> the customer system's jgroups configuration and
> started in the disaster recovery system.
>
> After that he recognized that initial_host was not correct. Because he used
> customer system's iniitial_host instead of disaster recovery system's ip.
>
> Customer's is
> <TCPPING timeout="3000"^M
>
> initial_hosts="${jgroups.tcpping.initial_hosts:100.254.163.31[7800],
> 100.254.16
> 3.32[7800]}"^M
>               port_range="2"^M
>               num_initial_members="4"/>^M
>
> After he changed the disaster's recovery system's initial_hosts from
> customer's to  disaster's system.
>
> *Problem occurred,* the problem is disaster's cluster joins the customer's
> cluster even disaster's inital_hosts is it's own ip, not a customers's ip.
>
> We were very difficult to  find this problem, and I reviewed whole jgroups
> sources and found some code.
>
> MERGE2 check the members periodically using TCPPING's dynamic_hosts.
>
> */** https://jira.jboss.org/jira/browse/JGRP-989 */*
> protected final BoundedList<PhysicalAddress>  dynamic_hosts=new
> BoundedList<PhysicalAddress>(max_dynamic_hosts);
>
> *When we have A,B and C listed in initial_hosts but later D and E join, and
> we have a partition {A,B,C} and {D,E}, when the partition heals later, we
> have no way of merging the 2 subgroups back into 1 because initial_hosts
> doesn't contain D or E.
> SOLUTION: dynamically add members not in initial_hosts to initial_hosts on
> view changes. Keep the number of dynamically added members bounded.*
>
> *This caused this problem, I think if the coord check the dynamic_hosts to
> merge other node, it also check the other node's initial_host*
> *whether it contains the coord's host address.*
> *Becuase D,E does not contains the ABC's IP address in initial_host anymore.
> *
> *
> *
> *And we could not restart ABC's node to remove the dnamic_hosts because this
> ABC are running in production mode.*
> *
> *
> *Thanks and i hope this help you to enhance my favorite jgroups.*
> *
> *
>
>
>
> ------------------------------------------------------------------------------
> Forrester Wave Report - Recovery time is now measured in hours and minutes
> not days. Key insights are discussed in the 2010 Forrester Wave Report as
> part of an in-depth evaluation of disaster recovery service providers.
> Forrester found the best-in-class provider in terms of services and vision.
> Read this report now!  http://p.sf.net/sfu/ibm-webcastpromo
>
>
>
> _______________________________________________
> Javagroups-development mailing list
>

-- 
Bela Ban
Lead JGroups / Clustering Team
JBoss

------------------------------------------------------------------------------
Benefiting from Server Virtualization: Beyond Initial Workload 
Consolidation -- Increasing the use of server virtualization is a top
priority.Virtualization can reduce costs, simplify management, and improve 
application availability and disaster protection. Learn more about boosting 
the value of server virtualization. http://p.sf.net/sfu/vmware-sfdev2dev
_______________________________________________
Javagroups-development mailing list