A new CTDB node grabs all public IP address in the cluster

[ronnie sahlberg]

You have two separate clusters, one with one node each.

/etc/ctdb/nodes must contain a list of ALL nodes in the cluster and
this file must be identical on all nodes in the cluster.


On Tue, Jul 26, 2011 at 12:16 AM, 曹宇 <caoyu at huaweisymantec.com> wrote:
>
> On nodeA, change the following files as following:
>  cat /etc/nodes
>  172.204.181.35
>
>  cat /etc/public_addresses
>  192.169.100.30/24
>  192.169.100.31/24
>
> On nodeB, change the following files as following:
>  cat /etc/nodes
>  172.204.181.36
>
>  cat /etc/public_addresses
>  192.169.100.30/24
>  192.169.100.31/24
>
> The do the following steps
> 1.execute "service ctdb start" on nodeA
> 2.change the content in /etc/nodes on nodeA to add a new node:
>   172.204.181.35
>   172.204.181.36
> 3.execute "service ctdb start" on nodeB
> 4.execute "ctdb reloadnodes" on nodeA
> 5.execute "ctdb ip" on nodeB, the output is:
>  Public IPs on node 1
>  192.169.100.30 node[1] active[bond0] available[bond0] configured[bond0]
>  192.169.100.31 node[1] active[bond0] available[bond0] configured[bond0]
>
>
> 于 2011-7-19 0:30, Jim McDonough 写道:
>>
>> 2011/7/17 曹宇<caoyu at huaweisymantec.com>:
>>>
>>> I found something weird.
>>>
>>> Start up a CTDB service on nodeA, which have the public ip
>>> of 192.169.100.20/16 and 192.169.100.21/16,
>>>
>>> Then start up a new CTDB nodeB,
>>> the result is that the new nodeB got two public ips(100.20 and 100.21),
>>> while nodeA got nothing.
>>>
>>> Why not nodeA and nodeB got one pulic IP each in this case?
>>
>> You probably need to provide more info here from ctdb status on each
>> node.  Are both ENABLED?
>>
>
>