Forwarding table, eh?
Cisco's Customer Engragement Feature rides again, eh?
Problems at Linode's data centre in Germany have led to connectivity issues with its cloud services over the weekend, and we understand the vendor is currently trying to solve a packet loss problem. At 14:14 UTC yesterday, the New Jersey-headquartered cloud provider said on its status page that it had become "aware of …
VMs are available from $5 per month.
Get two in different locations and mirror them. Nothing fancy, rsync and DB replication will do the trick for most web application servers. It's really straight forward and a simple DNS change saves your cured pork belly.
We go one step further and place the mirrored servers not only in a different location, but with a different service provider.
The golden rule : there is no such thing as 100% data centre availability.
> The golden rule : there is no such thing as 100% data centre availability.
But don't forget the secondary golden rules:
1. (Automatic) failover in the case of one host totally dying is easy. Failover in the case of sporadic application or network problems is much harder.
2. Recovering from split brain is very, very hard, and may amount to unrecoverable data loss.
In other words: unless your application is completely stateless, make sure your failover is controlled and manual.