* This post is an attempt to be helpful. I am not sure about any of it :) *
> I don't understand this explanation. How can
> there be a 'route open' to a dead node? And is
> the 'confirmation that the node went down' simply
> the loss of the TCP connection or something else?
I think that this is only refering to situations where you have two ethernet connections in some kind of bonding or HA setup - if you have two different interfaces set to different IPs I know that this is a not a problem. If you have two interfaces bonded together what it is saying is that it does not like failing over - and as far as it is concerned the node is "dead" because the first network interface died (and therefore the connection that was going on was not completed) yet there is still a route open via the backup interface. MySQL cluster does not like that, so do that at your own risk!
> b) tell me if there is any kind of workaround
What are you trying to achieve?
I can confirm that you *can* run a cluster just fine with multiple NICs as long as they are set to different IPs (typically a "private" and a "public" with the private on a reserved range such as 10.0.0.x or 192.168.x.x)
If you want more network redundency then simply run more nodes. You could run two seperate NICs and have two storage nodes on that box - with each using a different NIC allowing you some fault tollerance.
With best wishes,
Alex
>
> Thanks,
>
> Rick
Alex Davies
http://www.davz.net |
alex@davz.net