[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Linux-cluster] Fedora 13 Clustering



Hi,

Marc - A. Dahlhaus wrote:
Am Montag, den 12.07.2010, 15:06 +0100 schrieb Virginian:
  
Hi Gordon,

It looks to me like things have changed quite a bit in F13 compared to RHEL 
/ Centos 5. I agree, cluster.conf  looks the same but there is a new 
component called Corosync and ccs seems to have gone.
    

Take a look here:

http://people.redhat.com/ccaulfie/docs/Whither%20cman.pdf

>From RHEL5 to RHEL6 you change from cluster 2.X to cluster 3.X

Documentation can be found here:

http://sources.redhat.com/cluster/wiki/

  
Traditionally I have used Conga to configure the Cluster (although not 
without problems at times it has to be said). However, in F13 there is no 
"luci_admin" utility to set the password and Luci isn't accessible from a 
web browser. I don't mind configuring things by hand if that's the only way 
to do it, I was just looking to be pointed in the right direction really. I 
have received some help offline from a member of this list which has proved 
invaluable. I have now managed to get a basic cluster up and running on F13. 
Our fellow list member is working on a document which I think will be of 
enormous help to others when completed (in fact it is already a useful doc).
    

Docs on Conga (Luci and Ricci) are here:

http://sources.redhat.com/cluster/conga/

I don't know if this is newer because i made a large way around even
trying to use conga after i spotted that it utilizes turbogears (which
has a huge dependency tree just to get anything out of it).

  
The old docs specified which ports / protocols needed to be opened up in 
iptables but as I haven't sussed out what's needed under F13 I am running 
with iptables stopped for now. If anybody has this information I would be 
very grateful to receive it.
    

The FAQs on http://corosync.org/ and http://openais.org/ could help
here. Unfortunately the FAQ entry about firewalls has no content.

Bug?!

A "netstat -tulpen" on a root shell should show the ports that corosync
and openais listening on. They use a single multicast port if i'm not
mistaken.

  
In both corosync and openais, you define the ports and addresses it listens on. It's usually recommended that you have redundant links between the cluster nodes. Each link is represented by an interface declaration, in which you bind to a network card's IP address, you define a multicast group (might even require Protocol Independent Multicast routing to be enabled in some cases) and a multicast port. All nodes that should be on the same cluster should also communicate on the same multicast address and port, when using redundant links, both multicast addresses and ports must match, meaning each node will have 2 unique IP addresses, and 2 multicast addresses and ports shared between all nodes. Multicast is used for messaging and propagation of configuration changes. You can access any node and configure the cluster from it, changes are propagated automagically upon committing updates to the cluster resource manager, usually Pacemaker. Syntax is as follows.

        interface {
                ringnumber: 0
                bindnetaddr: 192.168.1.1
                mcastaddr: 239.1.1.1
                mcastport: 5405
        }

        interface {
                ringnumber: 1
                bindnetaddr: 192.168.1.2
                mcastaddr: 239.1.1.2
                mcastport: 5405
        }

Regards.

  
Thanks & regards

John
    

Marc

--
Linux-cluster mailing list
Linux-cluster redhat com
https://www.redhat.com/mailman/listinfo/linux-cluster
  

-- 
Dan FRINCU
Systems Engineer
CCNA, RHCE
Streamwide Romania

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]