There were two major outages on many sites during the past few weeks.
- Amazon AWS EBS degradation
- Outage due to Hurricane Sandy
Both these incidents affected many sites hosted in US East. Specifically, there were multiple sites that went offline during the recent Amazon AWS EBS degradation (nice alternative for downtime). Most sites waited helplessly until Amazon fixed it or waited until the after-effects of Sandy to go off. Others did the smartest thing. Moving the entire site to the US West coast.
So, what do you do?
The question in the title was asked in Quora. Kyle Hultman of Logicworks provided a great answer to mitigate this situation. Here’s the summary of it…
The best way to avoid failure from your web hosting provider is to use multiple providers. Failing that, use multiple geographic independent locations from your provider; Multiple availability zones. Layer on resilient DNS and load balancing to complete the offering.
At least this is what I do to prepare for the failure. Take urbancargo.com as an example. It runs on at least two servers at any time! Want to test it out? Use any of the following tools and type edge.neocha.com in the URL box…
The Plan
- Take regular backups of your site – The important thing to do after taking the backups is to make sure your backups work. So, take a latest backup and make sure it works. You may also want to choose a random backup and test it for its functionality.
- Have a working version of your site in another server – Hardware becomes very cheap and the hosting has become a commodity. You can choose a stand-by server at a fraction of cost than you pay for the primary hosting needs. Of course, your stand-by server should be scaleable in no time.
- Have a solid DNS manager – Stop using your registrar’s DNS service and choose a reliable DNS manager such as Dyn. Please also make sure you have a backup (secondard) DNS manager such as BuddyNS. In online, we never know when something goes down.
- Test it live – Shutdown your primary server and shut down your primary DNS. Test drive your backup plan, probably for a few hours on a low traffic day (weekend?).
Food for thoughts
Let me also share a couple of links to articles that provided some great insights on this…
