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

Re: [Linux-cluster] Ip Settings



Stewart -
Thanks for the info, this was exactly what I was looking for. I am working closing with our network engineer on this project, so I will be sharing this info with her and working out a solution.

B
On Jan 22, 2009, at 9:40 PM, Stewart Walters wrote:

burton simondsfamily com wrote:
I would like some advice on setting up a High Availability solution using RHCS with Apache.

I would like my clustered nodes to communicate on vlan, and the client connections to come through another vlan. Presently, I have 2 interfaces on each of 2 nodes. 1 interface on each vlan. I also have a vip address for the clients to connect to for the apache server.

node a:
xxx.xxx.100.1  eth0 (used for client connections)
xxx.xxx.200.1  eth1 (used for interconnect, cluster communication)

node b
xxx.xxx.100.2  eth0 (used for client connections)
xxx.xxx.200.2  eth1 (used for interconnect, cluster communication)

xxx.xxx.100.100 address for clients to connect to apache.


My question is how do I configure the apache service (and / or the resources) to use client interfaces using the vip address?

If i need to provide more information, please ask. This is my first go at setting up a cluster.

Thank you,
B

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


One way to do this is similar to a way that I have done on previous customer. We had several different NICs each with a VLANed IP and we used differing DNS names attached to each NIC to allow the traffic to be routed via particular NICs.

So you might want the following in DNS:

nodea.example.com = xxx.xxx.100.1
nodea-cluster.example.com = xxx.xxx.200.1
nodeb.example.com = xxx.xxx.100.2
nodeb-cluster.example.com = xxx.xxx.200.2

In the previous installation, we had to also put in place some advanced IP routing policies through the iproute2 package. The trick is to send any traffic received on the eth1 that back out via eth1 - not via the default route (which would be eth0).

Linux Journal has an article on how to do this titled "Overcoming Asymmetric Routing on Multi-Homed Servers" found at http://www.linuxjournal.com/article/7291 .

Then you just define in your cluster.conf that the two nodes are nodea-cluster.example.com and nodeb-cluster.example.com instead of nodea.example.com etc. etc.

We ran in to our fair share of network traffic problems with this sort of configuration due to the complexity of the network we were working with but once we had tuned it, it worked quite well.

If you can, I'd recommend you make yourself friendly with top-gun network engineer to help you assist you with networking issues should they arise. They can be invaluable when troubleshooting something that doesn't work :-)

Regards

Stewart

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


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