15 comments

[ 3.7 ms ] story [ 23.9 ms ] thread
You might consider multiple backup strategies, different types of backups done in completely different ways, none with any level of dependence on each other. Test them to see if you are able to restore from them.
Amazon pretty much provides that service for you. You just gotta turn it on.
If your only backups are on AWS, you better hope that your credit card doesn't start bouncing while you're on a week long vacation.
Be careful who you take advice from. I can't believe somebody who is providing information on setting up EC2 for startups was not using elastic block storage for their critical data.

There is really no excuse for this. Don't make the same mistake.

Yeah -- if I'm understanding that article correctly, he really had no business offering up advice on how to configure EC2 instances.

On the upside, the title is decidedly not link-bait. That really is the biggest mistake you can make on AWS.

I'm not entirely sure I see what he did wrong -- he says he was using 'block storage', but his remedy was to get an EBS volume?

Was he really running the contents of two websites on the EC2 instance?

I mean, I've done that. Once. When I was first playing with EC2, and well before I'd actually deployed a live and (apparently) valuable site to it.

If I'm understanding the problem correctly, then my sympathies. I don't know if it's a curse or a blessing that it went all this time without having rebooted and destroying everything before now. I suppose that's the sort of thing you'd prefer to learn earlier than later.

> Was he really running the contents of two websites on the EC2 instance?

Apparently so, despite the fact that all of the EC2 documentation labels the local disk as "ephemeral."

Is it "picking an instance too small to handle the traffic HN sends"?
I always picture the submit-and-boom scenario as skeet shooting, which keeps me amused enough until someone pastes a cache link into a comment.

Pull! Pow

Correct me if I'm wrong, but I thought the point of ec2 style clouds was to create an ami with everything configured just how you need it ahead of time, and boot lots of small machines off that. The data should be fed from s3 or from rds or from an ec2 instance with a database which stores it's data on an ebs.

The main idea is that it's an easy platform to scale horizontally from.

It never really struck me as intended to be a service which you run something like wordpress from...

What this guy probably wanted was something like rackspace and linode offer, which is a vm with a persistant disk, more in line with vmware on the desktop or esx, or xen, than amazons offering.

I agree, I think your development process on EC2 needs a strategy on how to regularly build new AMI images. Probably a mix of updating some things in-place while running, and sometimes creating a new baseline AMI image of your whole OS and app, which is then used on the next reboot.

Also I would store all dynamic data on RDS (database) and S3 (uploaded files), probably no need for EBS in most web apps.