Le vendredi 03 septembre, Jangita écrivit : > Hi List, > > This is probably not the list for this question, but I guess it is a > good starting point since I used debian servers for all my > deployments. > > I want a scenario where I have two servers that appear as one server > (one IP) to the outside world, with both servers having *exactly* > the same data and services. So in case one server falls into a pond > somehow, I woun't even loose a ping from my clients. Prepare to loose some pings !!! There is only some specialised harware that can be fully redondant with a really transparent failover (Tandem systems use to be one on them). Nowadays, I know that we can have a similar sttings with VMware Virtual Machines, running in parallel, synchronized. But, as the machine must exchange informations like time, random generator, and frequantly resynchronise thier memories, it is not much performant. The real pain for HA is data. Applications can be switched fast and easily, especially Web applications. Either you have a SAN/NAS wth two Arrays and synchronous replication ($$$), or you use DRBD. There is some new distributed filesystems or storage systems who emerge : - Ceph - glusterfs More specific for Virtual Machines : sheepdog Or key-value : all the No-SQL movement, redis, cassandra...
Description: Digital signature