Follow-up August London/Frankfurt outage questions
I appreciate the detailed postmortem about the August 1-2 London/Frankfurt outages but I have further questions.
Why does a network outage in London also cause a network outage in Frankfurt?
Is Linode making changes to reduce the risk, impact or duration of such outages?
Do other data centers have the same circumstances?
How should a customer design an HA setup using multiple data centers?
<3
1 Reply
✓ Best Answer
To address your questions, I've included my answers inline below:
Why does a network outage in London also cause a network outage in Frankfurt?
This was related to networking configuration where we were trying to set up more economically efficient routing. When it worked, it was good, though this incident was a reminder of how bad this was when things didn't work.
Is Linode making changes to reduce the risk, impact or duration of such outages?
After that incident, we implimented changes to prevent this sort of issue affecting multiple data centers.
Do other data centers have the same circumstances?
London and Frankfurt were the only data centers with this sort of circumstance. Now with the implemented changes, there are no data centers like this.
How should a customer design an HA setup using multiple data centers?
The best way to set up geographic redundancy would be through a Content Delivery Network (CDN) service. Akamai has been offering a CDN service for a while: Global Traffic Management - Load Balancing Solution | Akamai