Paul, Thanks for pointing that out. The marketing page is confusing and we will correct it.
Our postgresql database is not replicated in the sense that you would expect (i.e. hot/cold standby for failover). It does however use a replicated EBS (http://aws.amazon.com/ebs/) RAID array, which provides significantly enhanced data-durability. Historically this has resulted in almost no incidents of data loss for Heroku app. If a replicated database is a requirement, you can still achieve this with Heroku by using Amazon's RDS database (http://aws.amazon.com/ rds/), and our free RDS add-on (http://addons.heroku.com/amazon_rds). Best, Matt Soldo On Nov 17, 10:03 am, Paul Dowman <[email protected]> wrote: > Thanks for the reply. So it's not actually possible to guarantee that > there will be no data loss, the best we can do is an hourly backup > (assuming the data set is small enough that a full dump each hour is > feasible). > > So why does the marketing page (http://heroku.com/how/architecture) > say that there's replication? > > Thanks, > Paul > > On Nov 16, 5:02 pm, Peter van Hardenberg <[email protected]> wrote: > > > > > > > > > Hey Paul, > > > sorry -- I'm super-busy right now but I'll at least tap out a bit of a > > reply. > > > On Tue, Nov 16, 2010 at 1:34 PM, Paul Dowman <[email protected]> wrote: > > > Hey Heroku guys, just bumping this thread. > > > > To summarize: do we need to do automated regular backups to protect > > > against Postgres or some other part of Heroku infrastructure going > > > down, or is the database guaranteed to be reliable? > > > We take automated backups as disaster insurance, but make no promises about > > their intervals. In the event of an outage, we handle recovery. If there is > > the potential for data loss, we reach out to any affected users. > > > > I'm guessing we do, and if so how do we do that since an hourly dump > > > of postgres via cron isn't reliable enough or scalable? (i.e. you can > > > lose up to an hour of data, and more as the dump starts to take longer > > > with a large dataset.) > > > Hourly dumps is probably your best solution at the moment, but we're aware > > that there are better solutions out there and would love to schedule those > > into our release schedule some time soon. > > > Having said that, in the three years we've been running PostgreSQL, I > > believe the number of data-loss failures (and by that I mean > > restore-from-backup failures) could be counted on one hand. > > > -pvh -- You received this message because you are subscribed to the Google Groups "Heroku" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/heroku?hl=en.
