I need to setup some servers for redundancy (same data center). I saw this now: http://www.howtoforge.com/high_availability_nfs_drbd_heartbeat I will use Debian and Cpanel for this project. My questions regarding that article: - what exactly is the "virtual IP"? For example: Let's say the resolvers (yes, resolvers in resolvers.conf) 192.104.107.1 147.202.2.54 The *server* IP for server #1 is xxxxx.66.2 The *server* IP for server #2 is: xxxxx.66.3 What/where is this "virtual IP" the article refers to? I'll admit I scanned the article quickly, so bear with me Also, any issues using cpanel / debian for this method? Partner's outlined architecture / setup goal: box 1 – site #1 - some sort of redundancy - local backup of mysqldbs, and web site files box 2 - media server - Debian OS - local backup to NAS with rsync (about 80GB but growing) media server (ftp) - not sure where that is now - but seems like a good place to put it ability to easily add ftp access to specific directories for some customers to upload audio content staff ftp sharing area public ftp area that staff can upload to box 3 - - centos - redundant server setup
I guess you have a subnet at your hoster where the IP addresses xxxxx.66.2 and xxxxx.66.3 are from. Pick an unused IP address from your subnet (e.g. xxxxx.66.4) and use it as the virtual IP address (this works only if your hoster doesn't bind IP addresses to MAC addresses so you should ask them). The virtual IP address floats between the load balancer, i.e., normally it'S assigned to the active load balancer, but if that load balancer goes down, the passive one takes over the virtual IP.
Ok, which load balancer are you referring to? My idea was using Heartbeat and DRBD on two servers, one master and one slave. Which load balancer are you referring to here? Thanks
I guess the loadbalancer is here meant as the "active node". That means just that your clients access always one IP-Adress (the virtual one). Depending on the hearbeat status, the active machine will take ownership of this IP. That's the normal way an active/passive cluster works.
Ok, the setup with HB and DRBD I want is this: Main Server (Master), Box 1 Mirrored Server (Slave), Box 2 So this is correct? DRBD/HB is to be installed on both servers according to the tutorial.
Another question: If box 1 goes down, why would box 2 *not* go down if both are mirrored? So if box 1 gets slashdotted (whatever the term is) and can't handle the load, we'd expect the identical mirrored server to likely crash too, correct? Same idea with misconfiguration on box 1, it would pass on two server 2. So where is the advantage?
