I was wondering if someone of you can share some experience with FreeBSD HA clusters. What are the best practices and what software can be used to make two servers with shared storage highly available ?
Thanks for the reply. As far as I know HAST is used to store data on two physically separated machines and CARP is for IP sharing between two nodes. That is not what I am looking for. I would like to know if there is something like Service Guard, Sun Cluster, or HACMP that works on FreeBSD. In other words: when one node fails, the applications running on it to start automatically on the fail over node.
Perhaps you could look into it? Doesn't seem like much use to ask a question and dismiss the answer on assumptions when you have no good reason for those assumptions. ucarp does exactly what you are talking(see ucarp_up.sh and ucarp_down.sh) about and many people need HA storage as well which is HAST. Heartbeat is also available and possibly others.
Thanks Galactic. I suppose I was wrong with my assumptions. I will take a deeper look and see if carp and hast can suite my needs. However, probably instead of making remarks, you can just point me to some docs to read and share some experience. I think that this is what this forum is all about.
I'm not aware of any direct equivalents of HA clustering that runs on FreeBSD either.
The HAST/CARP solution may work for you, but while having the advantage of not requiring dual attached (often expensive fibre channel) storage from reading about this seems to have the disadvantages of degraded disk IO because we are replicating all disk updates via TCP/IP. If you can test the IO performance and its sufficient then it seems a neat solution.
Otherwise you might have to look at Linux and LinuxHA (or Veritas cluster if you can pay).
Id be interested to know of anyone actually usuing HAST and CARP for this type of configuration. Is this reliable under FreeBSD? What about network issues, I imagine both HAST nodes could try and run as masters where network connectivity is down between them, what happens when network connectivity is re-established? Thats something HA solutions typically deal with in one way or another (ie via multiple redundant network heartbeats and SCSI fencing etc).