On Fri, 20 Aug 2004, Andrew Piskorski wrote:

On Fri, Aug 20, 2004 at 08:45:06AM -0600, Ara.T.Howard wrote:
On Fri, 20 Aug 2004, Matt Sergeant wrote:

As far as linux -> nfs access goes it all depends on the quality of
the NFS implementation, and the kernel drivers you're using. It's
not easy to create a stable NFS locking system. We've had lots of
problems with it. If you can, go with local disk.

i cannot - the system is a priority queue system where the queue db is accessed via around 30 nodes. i am wrapping any code that will grab a write lock with an nfs safe lockfile creation to ensure only

So, why are using embedded SQLite code on each of the 30 nodes to access the single database file over NFS? Isn't your situation EXACTLY what the client-server RDBMS was invented for?

Why don't you simply run PostgreSQL (or your own little custom RDBMS
using SQLite) on one machine, and have all 30 clients talk to it over
the network?

we're a dot.gov and, since sept. 11th our host a clamped down to an unbelieveable level - ssh only. period. nothing gets in or out except on 22. additionally this system must be 24/7 and setting up HA setups for postgresql is tricky (i've done it twice before). i'm interested in the new mysql cluster for this, but it's pretty new. the thing is, all processing stops w/o this particular nfs server (it's where all the code lives) so it's already a single point of failure - i simply don't want to ADD a point of failure. your point, however, is well taken. the other problem is the security involved with connecting to and rdbms with code running all over the place - our security rules are totally dranconian and even this is a daunting task. nfs is the last refuge! ;-)

-a
--
===============================================================================
| EMAIL :: Ara [dot] T [dot] Howard [at] noaa [dot] gov
| PHONE :: 303.497.6469
| A flower falls, even though we love it;
| and a weed grows, even though we do not love it. | --Dogen
===============================================================================

Reply via email to