|
I may be wrong but it sounds like you are under a misconception here.
Using Piranha here is how things work.
You have 2 front end nodes (call them vnode0 and vnode1)
You have 3 web servers (call them ws-00, ws-01, and ws-02) these are the
real servers.
On the front end nodes (vnode0 and vnode1) you have piranha running which
consists of nanny, pulse, and lvs.
You have nothing other than apache running on the real servers.
In the /etc/lvs.cf file on vnodes 1 and 2, you define the following
backup_active = 1
backup = vnode1_ip
heartbeat = 1
heartbeat_port = 539
This tells the lvs software that when you start up pulse it is to check for
a backup server (which is the vnode1 in the configuration). It tells it that
your backup IS active and that it is to go into a heartbeat configuration
listening on port 539 and talking on the same to the backup node (vnode1)
Now, what FOS is in this situation is that if vnode0 should go down for
whatever reason, if you have pulse running on the backup node it will take
over (this is your failover) serving the VIP and handling the routing
table(s) for folks coming in on the VIP to access the real servers.
Now you can also run FOS mode which is the service= line. In THIS
configuration you ALSO have a backup node BUT the key differences are as
foolows.
All your services are running on vnode0, the same thing on vnode1. If vnode0
dies for some reason vnode1 will take over. There are no real servers like
there are in a cluster formation (lvs). ERverything is running on the
primary and the backup node. The backup node (vnode1) stays dormant save for
the fact that it talks back and forth with the primary node (vnode0) waiting
to see if vnode0 fails. As soon as it sees that vnode0 is not responding it
waits the specified time in the config file before IT assumes the role of
vnode0.
I hope this makes things a little clearer.
>
>Dear John Cronin,
>
>> > I have setup a primary node, a backup node and two
>> > real servers for my LVS. However, if not mistaken, I
>> > know that piranha canot handle lvs and failover at the
>> > same time. Is there any solutions to make lvs and
>> > failover possible at the same time?
>>
>> I am not clear what you are trying to accomplish here. If you
>> want failover between the primary LVS directory, and the backup,
>> my impression is that Piranha should in fact handle that properly.
>> Otherwise, what would be the point of having a backup director?
>>
>> I think it is in the realservers where you can have load balancing,
>> or failover, but not both. That is moot however, since with nanny
>> and a loadbalancing Piranha setup, failed servers are supposed
>> to be removed from the routing table, which is failover as far
>> as a load balancing service is concerned. Failover is just
>> a way of using a third party (the director) to heartbeat two
>> services and control failover between them.
>
>What I mean is when I configure the failover between primary and backup
>node using piranha, it is doing failover only. The primary or the backup
>(when primary is fail) node didn't do the load balancing to the 2 real
>servers. Is there any ways that I want to have failover between the
>primary and backup node, and at the same time, when requests come in,
>load balancing is functioning. (The primary or the backup will load
>balance the requests to the real servers)
>
>
>
>Thank you
>
>
|
