Author Topic: Sequoia Database High Availability  (Read 2339 times)

hans

  • Guitar Addict
  • Jackass In Charge
  • Posts: 3523
  • Karma: +46/-18
Sequoia Database High Availability
« on: May 11, 2007, 01:02:26 AM »
I just got a fail over/load balanced database solution figured out using Sequoia (C-JDBC). We're trying to put together a high availability solution where I work so we won't have to worry about when we decide to do maintenance (web layer too). It took a bit of messing around to get it working correctly in the scenario we're trying to put together but know that I sort of understand how it works, it's actually pretty simple to do. I'm still working on figuring out how to automate the recovery but hey I've only spent a day so far on it.

So either one of the of the Sequoia controllers/DB's could go down and the other one will simple take over. It's actually a really cool architecture. In theory you could run everything on two boxes and have a mirrored/load balanced setup (there's a load balancer in front of the web servers too).


Right now it looks something like this (with a web load balancer in front):
Code: [Select]

    Web1           Web2
      \             /
       \           /
      Sequoia JDBC URL
             |
             |
             V
Sequoia1  <----->  Sequoia2
   |                  |
   |                  |
  Db1                Db2

This signature intentionally left blank.

ober

  • Ashton Shagger
  • Ass Wipe
  • Posts: 14305
  • Karma: +73/-790
  • mini-ober is taking over
    • Windy Hill Web Solutions
Re: Sequoia Database High Availability
« Reply #1 on: May 18, 2007, 12:00:30 PM »
Nice... we recently expanded our architecture where I work.  We now have 3 application servers but still only one DB server.  With the nature of our business, we can't duplicate servers in production and run them concurrently.  The best we can do is mirror one to the other so if one goes down we can switch to the backup.

The good news is that we never have to take the application down to update it... they've even managed to figure out how to roll sessions from one app server to the next so people never notice if a server goes down or if we have to deploy something.