Desired: DR that meets the business objectives (even if it sounds moronic)
1. Three critical web-based internet apps that need to failover.
2. No automatic failover. Management wants to decide at point of failure the ROI of whether to failover or fix the failed app.
3. Failover is per application (planned or unplanned). You could have two apps running from the Datacenter, but the third one running from the DR.
4. Of course, in case of total site outage, we need all apps to failover to the DR
5. We have a /16 address space. One app use CSS?s for loadbalancing using VIPs; the others use clustering (using IBM Domino and Websphere).
6. We happen to have two GSS?s sitting idly for a year.
7. Under normal conditions, the three applications are up and live on both DC and DR, although the apps on the DC are the ones that get exposed to the Internet.
1. DR site uses different IP space. No BGP AS prepending or conditional advertisement kind of game. Instead, use GSS to do name resolution.
2. Use GSS as authoritative Name server for the three apps.
3. To simulate manual failover (business requirement) for apps/vips, we propose to use the ?ordered list? balance method but with a twist: GSS will check the health of an app at the DC, if it detects failure, normally, it will automatically go to the next item on the ordered list, which would be to look at the corresponding app running on the DR. But we want to simulate manual failover. So, we let the GSS have a duplicate entry both pointing to the same app running at the Datacenter. GSS checks the first entry. Finds that app has failed, goes, according to the ordered list method, to the next entry which still points to the same app and thus will register (again) failure; at which point we are hoping the GSS will do the ?right? thing, namely, ?gracefully? hang while in parallel sending application administrators an email or an alert indicating that manually intervention is needed. If a decision is made to failover this particular app; then make a manual config change on the GSS to have a duplicate entry in the GSS answer group similar to the above but in this case pointing to the DR, rather than the Datacenter.
4. My sincere apologies if I confuse anyone, but I am new to GSS and to DR design. Am I way offbase? Can the GSS do as above? Can it also send alert or notification on application failure? If the above has weakness, can you please suggest improvement or alternative.
5. Your input greatly appreciated!!