SimpleDB gotcha: they can delete your data | Hacker News Reader

SimpleDB gotcha: they can delete your data

dailyawswtf.com

I think this is just so that Amazon can avoid permanently storing data from idle accounts which are within the free-storage quota. It really ought to be better clarified, though.

Maybe so, but the obvious solution then is to give users an option to disable the free portion of the service. Disappearing data ain't worth the $25 a month or whatever you save with the free usage.

So long as I'm paying for the service and storage, stay the hell away from my data!

Perhaps, but it's pure laziness. Amazon should simply archive the data if they want to free up capacity on their SimpleDB servers. I can't imagine SimpleDB's servers are struggling under the load of their dozen-or-so users.

Eventually consistent, and even then only temporarily so.

This is really unfortunate since it prevents using SimpleDB as the system of record for any infrequently accessed data.

Well it isn't a guarantee that they will delete the data, just a notice that they won't be liable legally if they do decide to do so. Probably just a legalistic CYA, but if so they should probably word it better.

I agree it's probably just boilerplate that they're unlike to exercise, but I do think it's part of the SLA and so it's entirely appropriate to single it out and ask for clarification or change.

"Well it isn't a guarantee that they will delete the data, just a notice that they won't be liable legally if they do decide to do so."

That's even worse. If I know my data will be deleted, and under what circumstances, I can take steps to address the issue. "We might, potentially delete data, probably not, but who knows?" is kind of hard to plan against.

Loss of data after 6 months did not seem that bad after I read that Amazon shuts down your MySQL (RDS) instance every week for up to 4 hours "for maintenance"

http://aws.typepad.com/aws/2009/10/introducing-rds-the-amazo...

Edit: wording fix

You should read it again; there's a 4-hour "maintenance window" every week where any required changes will be applied, but downtime is limited to the time necessary to apply the changes, not the entire four hours. And you can select the 4-hour window that's most convenient for you.

For larger sites, this seems like it would hardly be a problem...if you have a pool of RDS instances, you could just stagger the maintenance windows so you never have more than 1 of your instances down at a time.

Why not just write a cron job to insert a row every now and then eh?

The wording of the policy indicates that they can selectively delete just the data which has not changed.

The policy says accessed, not modified.

Because that's a pretty messy solution to what should be a non-problem. You shouldn't have to "refresh" the data in your database every now and then, just like you don't need to buy a new key for your car every couple weeks.

Reply on news.ycombinator.com