[GE users] bdb server for multiple clusters?
davidr at ressman.org
Mon Aug 2 13:51:07 BST 2010
For various reasons too painful to get into, I'm running a set of independent GE clusters for various groups at my office. The gridengine servers run in KVM virtual machines, all the primary hosts are on one physical server, and what will be their shadow masters are all running in VMs on the backup server. We have another server with
very fast disk that's running the arco DB (among others). This seems to me like a natural place to run the BDB spooling database for the gridengine instances.
That said, from looking into the installer, it looks like setting the one server up to act as the spooling db server for all of the
gridengine clusters would be rather difficult. I'm not even sure if it's possible, since, even if I took the time to hack up all the startup scripts, I don't know how portmapper would react to this. All I see is that the db server registers program "351457", which is sleepycat... er.. our New Oracle Overlords' bdb server number. If I
tried to start up more than one of these, would it work at all? I would think that the RPC clients wouldn't know how to ask for the
proper instance if portmapper even let me register multiple servers.
If not, what is the best way to use one host hold the data needed for failover from multiple gridengine servers? I worry about using a common NFS filesystem to hold the spool because some of our groups will be spooling tens or hundreds of thousands of jobs in a day.
To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].
More information about the gridengine-users