Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I'm now using AWS RDS multi-AZ with great results, the problem is that I want my infrastructure fault tolerant between regions. I'm looking for some kind of tool to make a cluster with RDS and my servers.

As I read you cannot cluster directly from RDS to another server, just Read Replicas, but not between regions neither outside AWS.

I though about do some kind of mysqldump every 15min or something like this but this is not efficent, and the restore could be hard to maintain.

In your experience, what is the best thing to do, run my mysql myself and configure clustering with my machines, or do something else. In this way, there are some kind of cold replication that doesn't affect performance? I mean, every 5 minutes sync changes. I can afford the lost of a few data but not the downtime ;)

Any ideas, would be apre

share|improve this question
    
Do you need to use a MySQL database (AWS RDS is MySQL)? If not, you can look at some other datastores that are geared around automatic cross-az/cross-region replication. –  yfeldblum Dec 2 '11 at 12:47
    
do you mean simpleDb? I need relational db, and everything is developed with mysql, I don't really know which I will have to do with a change os db engine. –  enedebe Dec 2 '11 at 14:11
add comment

2 Answers

Configuring your own MySQL cluster (with replication in multi regions) will be time consuming, costly, and require expert advice to help setup and maintain. Having said that it can be more reliable than the current RDS solution (only if you've done it properly) as it could survive a whole region failure.

If you really need 100% uptime then rolling your own solution is an option you should look at. Do you have the money, time and resources to do so? Can you afford to invest the thousands and thousands of dollars or would it be cheaper to possibly have a little bit of down time?

RDS has suffered issues before (as recently as 2012). Interestingly the following site reports that this risk could be greatly minimised by using the "Point-in-Time Restore option"...

http://www.networkworld.com/news/2012/102312-amazon-outage-263617.html

Like with the EBS issue, AWS reminded customers that if they enabled Point-in-Time Restore option, then they could launch a new database instances using a backup of the impacted database in another availability zone.

Another option could be to simply make sure you have access to your backups. For example if you have daily backups you might send those to S3 for easy access (or maybe outside Amazon once a week if you're paranoid). In the event of an EBS/RDS failure you might be able to create a new RDS instance and restore it quicker from S3 in the case that RDS has some serious issues. This solution assumes you're okay with a few hours of downtime in favor of not having to do lots of work engineering some crazy cross region solution.

Finally, depending on your application it might be cheaper to try and take advantage of an non-relational database. I know you've stated that you need a relational DB, however re-engineering some or all of your application to use a non-relational DB may be easier & cheaper then rolling your own MySQL across multi-regions (non-relational also give you other benefits around scale etc). This solution isn't for everyone, and may be in the too hard basket for you (however future projects might be easier to implement this way!).

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.