Hello:I am trying to implement a high availability cluster with 2 nodes with centos 6.3, if i shut down the server one the server two takes the virtual ip address and the services are ... [by tania]
I'm in the process of configuring a sort-of Beowulf cluster, the major difference being that I want to be able to add nodes from any network. Because of this I can't run DHCP on the server to inventory all the nodes.
New nodes would run a package or script to configure SSH, MPI, NFS and the like. During installation they'll require the IP address of the server.
I'd like to build a two or three node Pacemaker cluster that would manage Xen (or possibly KVM) virtual machines.
Is it possible to set up the Pacemaker cluster (or some of its components) in such a way that when all nodes are active the VMs will use all available RAM but when one node fails the cluster will reduce VM's memory to fit them all on remaining nodes?
I have a 10 node cluster and multiple clustered resources on the cluster have IP Addresses that move between different servers in the cluster when the application is migrated.
This causes RSYNC and other tools that use SSH to fail because from the remote server the key it has in known_hosts for this virtual server no longer matches the one used by the node that is now hosting the application.
After three days of intensive Google searching, I have not found any high-level workflow of how to build a low profile - cheap - computing cluster (we are not interested in HA yet). This is just a front-end plus a node for now.
We want to start small with rockscluster, provide a web-based server for offering services, and then add nodes as our budget increases.