[Linux-cluster] Cman doesn't realize the failed node

[Hakan VELIOGLU]

Hi,

I solved my problem. When the kernel IP forwarding feature  
(/proc/sys/net/ipv4/ip_forward) is 0, then cluster nodes don't realize  
the failure. I write this solution to help others.

However, I am curious about that is all of your RedHat 5 OS default  
ip_forward settting is enabled? Are all your failover clusters working  
as expected ?

Have a nice day list.

PS: This change is included just in Red Hat 4 Cluster Suite  
documentation not in Red Hat 5 cluster suite. Interesting!!!

----- veliogluh at itu.edu.tr den ileti ---------
   Tarih: Wed, 12 Nov 2008 13:17:00 +0200
  Kimden:  Hakan VELIOGLU <veliogluh at itu.edu.tr>
Yanıt Adresi:linux clustering <linux-cluster at redhat.com>
    Konu: [Linux-cluster] Cman doesn't realize the failed node
    Kime: linux clustering <linux-cluster at redhat.com>


> Hi,
>
> I am testing and trying to understand the cluster environment. I ve  
> built a two node cluster system without any service (Red Hat EL 5.2  
> x64). I run the cman and rgmanager services succesfully and then  
> poweroff one node suddenly. After thsi I excpect that the other node  
> realize this failure and take up all the resources however running  
> node doesn't realize this failure. I use "cman_tool nodes" and  
> "clustat" commands and they say the failed node is active and  
> online. What am i missing? Why cman doesn't realize the failure?
>
> [root at cl1 ~]# cat /etc/cluster/cluster.conf
> <?xml version="1.0" ?>
> <cluster alias="kume" config_version="54" name="kume">
>         <totem token="1000" hold="100"/>
>         <fence_daemon post_fail_delay="0" post_join_delay="3"/>
>         <clusternodes>
>                 <clusternode name="cl2.cc.itu.edu.tr" nodeid="1" votes="1">
>                         <fence/>
>                 </clusternode>
>                 <clusternode name="cl1.cc.itu.edu.tr" nodeid="2" votes="1">
>                         <fence/>
>                 </clusternode>
>         </clusternodes>
>         <cman expected_votes="1" two_node="1"/>
>         <fencedevices/>
>         <rm>
>                 <failoverdomains>
>                         <failoverdomain name="domain" ordered="1"  
> restricted="1">
>                                 <failoverdomainnode  
> name="cl2.cc.itu.edu.tr" priority="1"/>
>                                 <failoverdomainnode  
> name="cl1.cc.itu.edu.tr" priority="2"/>
>                         </failoverdomain>
>                 </failoverdomains>
>                 <resources/>
>                 <service autostart="0" domain="domain"  
> name="veritabani" recovery="restart"/>
>         </rm>
> </cluster>
> [root at cl1 ~]#
>
>
> When the node gows down, the TOTEM repeastedly logs messages like this.
> Nov 12 13:12:57 cl1 openais[5809]: [TOTEM] The consensus timeout expired.
> Nov 12 13:12:57 cl1 openais[5809]: [TOTEM] entering GATHER state from 3.
> Nov 12 13:13:03 cl1 openais[5809]: [TOTEM] The consensus timeout expired.
> Nov 12 13:13:03 cl1 openais[5809]: [TOTEM] entering GATHER state from 3.
> Nov 12 13:13:09 cl1 openais[5809]: [TOTEM] The consensus timeout expired.
> Nov 12 13:13:09 cl1 openais[5809]: [TOTEM] entering GATHER state from 3.
> Nov 12 13:13:14 cl1 openais[5809]: [TOTEM] The consensus timeout expired.
> Nov 12 13:13:14 cl1 openais[5809]: [TOTEM] entering GATHER state from 3.
>
>
>
> Hakan
>
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>


----- veliogluh at itu.edu.tr den iletiyi bitir -----