[Linux-cluster] porblem with quorum at cluster boot

[jacob ishak]

hi
in 2-node cluster you don't need a quorum disk , there is a 2-node cluster
option that will added to the cluster.conf , and the cluster will be
functional quorate and everything , think of it  if you have 2 nodes only .

regarding your problem , is the problem occuring on a specific node , or on
both , if on a specific node , then check connectivity with shared disk ,
also how are you shutting down the cluster ( hard shutdown? ) ,
simultaneously or at the same time, .... it all maters ?? also check the
fencing devices is the fencing occurring correctly ( did you tried to unplug
network cables from one node and observed behavior )???? many parameters to
check , my recommendation is to troubleshoot step by step starting with the
basics ( hardware .. ) and on the way up .

On Fri, Sep 24, 2010 at 7:38 PM, emilio brambilla <emilio at inet.it> wrote:

> hello,
>
> I have a 2 node cluster with qdisk quorum partition;
>
> each node has 1 vote and the qdisk has 1 vote too; in cluster.conf I have
> this explicit declaration:
> <cman expected_votes="3" two_node="0"\>
>
> when I have both 2 nodes active cman_tool status tell me this:
>
> Version: 6.1.0
> Nodes: 2
> Expected votes: 3
> Quorum device votes: 1
> Total votes: 3
> Node votes: 1
> Quorum: 2
>
> then, if I power off a node these value, as expected, changed this way:
> Nodes: 1
> Total votes: 2
>
> and the cluster is still quorate and functional.
>
> the problem is if I power off both the node and them power on only one of
> them: in this case the single node does not quorate and the cluster does not
> start: I have to power on both the node to have the cluster (and services on
> the cluster) working.
>
> I'd like the cluster can work (and boot) even with a single node (ie, if
> one of the node has hw failure and is down I still want to be able to reboot
> the working node and have it booting correctly the cluster)
>
> any hints? (thank's for reading all this)
>
> --
> bye,
> emilio
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20100925/e6d21735/attachment.htm>