Maintenance Downtime on Tue. 3/6/12 4 pm UTC

As described in the last blog post, we haven't been able to perform the actual server upgrade last Saturday. Now we've chosen to go a safer way. We prepare everything, and will simply point spaaze.com to the new server when ready. Since Spaaze isn't using a Load Balancer yet (I can only hope that there will be so many people in the future that we'll have to scale our infrastructure) I prefer to turn off the old one while the transition is in progress (to not mix up states between the two servers). This will be much less painful than what I had planned last Saturday, and need less time. So hopefully tomorrow evening there is only a short downtime while the DNS updates.

Update 1 (4pm): Maintenance starts now...
Update 2 (5:30pm): The maintenance finished (although it again took longer than expected). spaaze.com now points to a new IP address, so it could take some more minutes until Spaaze is really reachable again and this IP change propagated through the domain name system.

Debriefing:
Overall the server switch went well. Although the transfer of the current state from the old to the new server took longer than expected, once copied, all services of the server worked fine, and the propagation of the DNS change was only a matter of time. One thing did go wrong though and was just corrected now (15 hours after finishing the server move): The whole idea of the state transfer (the thing that took so long), was to keep all users logged in, even on the new server. Due to a config error that did not really work out. I just corrected this now, and it should behave as if nothing happened. Unfortunately, all users who logged in within the last 15 hours may have to log in again - and if they were online during the config change maybe lost their connection to Spaaze. Sorry for that.

Comments:

Good Idea

Hey Guys/Gals... I think that is a very wise course of action. It is almost always better to do a migration then an in place upgrade... much better fallback options. Thanks!