For the Fedora deployment, our infrastructure team wants to run Askbot in multiple machines for redundancy/high availability and have those connect to a single DB (since our DB system is not replicated yet). Is there any problems with this setup?
We are using Postgresql 8.4 with python-psycopg2-2.0.13 in RHEL 6
At the level we were asking, it's a coding question ;-) most web applications these days have an architecture where the only thing shared between instances is the database (Used to be called Shared Nothing http://en.wikipedia.org/wiki/Shared_nothing but the database is shared so it's not really.) This allows deploying the web application to multiple application servers easily and the database is the only SPOF.
There are, however, still some apps that are coded that need to share other resources besides a database. For instance, we ran an application that needed to cache code repositories from hundreds of projects around the internet. This was an expensive operation in terms of disk and bandwidth and the cache could get out of sync between app servers. We ended up needing to create a nfs share for this data and forcing the cache to be created and synced only on a single node.
So really, we're asking whether askbot is a standard web app that just needs to share the database or if there's other dynamic, shared resources that we need to know about.
I think this is fully a deployment issue, hopefully your system administrators can figure this out. If you think something is missing in askbot to support this - please let us know.
Comments