Thanks a lot. This article helped me a lot. Just an nfo:
The mysql root password is rootPASSWORD (For the ones who couldn’t find or couldn’t reset (like me) )

Cheers..

I think your fears are misplaced.

We’ve been running our web servers and VoIP exchange on EC2 for 2 years without a hitch - this blog is on a server hosted on EC2. The mail server has been running for a month now. We started the mail server using spot pricing and expect that, sooner or later, it will switch of and we’ll need to start it again.

But that’s our choice. We’ve chosen to take the risk that on a particularly busy day spot prices may spike and we’ll need to restart this server. At that time we’ll decide whether to continue with spot pricing or pay the ‘retail’ price - as we do with our web servers - when the service is not affected by price.

No hosted service is 100% reliable and I’m sure we’ll face some problems but we’ve not faced them in 2 years. By contrast our in-house servers are much less reliable.

I’ve heard about AWS outages but I think they’re a bit of an urban myth or may one that affects S3 more than EC2 because, as written, these outages have not affected us.

Of course there have been times when one or more of the web servers have been unavailable but these issue s have always problems of our own making (for example by screwing with the OS)

I can only report what I know and I hope you do move to EC2 and enjoy the benefits we have. We’ve been able to cut our hardware budget *and* have a more reliable service with better backup.

Bill

I have people that rely on that 24/7 so EC2 seamed to be interesting with the scaling part but the fact that an istance can be brought down at any moment it scares me a bit.

thanks for the article, i,ll poke around for that and maybe transfer some of my personal stuff on there.

However we’re using this as a public relay server. All email is checked and sent directly to our in-house server over a secure connection.

Yes, if we were using it as a mailbox store (which the server is capable of being) we’d have to store the mailbox folders on something like as EBS device as you suggest.

I’m no mail expert, but isn’t it possible that you could lose mail with this strategy? Or is the EC2 server just an SMTP server, and not an IMAP server?

IF it was an IMAP server, couldn’t you lose drafts of messages, or have a message delivered successfully, then have failure and lose the message?

Again, I’m not a mail expert, but both of those seem like possibilities.

What am I missing?

Thanks.
Dan

For now we’ve just bundled the image so in effect, yes, for the moment we’re persisting to S3.

ami-395fba50 is based on Aelastics Ubuntu Hardy and predates the ability to boot from EBS so that one’s out - at least until there’s an Ubuntu based on one of the newer kernels and we can find the time to create a new machine.

I the mid-term we’ll do what we do for our web site: add an EBS volume and move the log files there.

However in the event of a failure its only the mail log files we’d lose so we’re not rushing to change anything.

Our email service used to be run by a company called Easyspace and these servers are still there (it was free) - just with lower MX record priorities. So in the event of a failure, mail will still be collected because senders will pick the next server in the MX list.

Sooner of later we’ll realise the issue and reboot from our custom image.

Great article. One thing I didn’t notice was any mention of EBS. Do you have your mail directories backed up to another server, or are you persisting it to S3?

I’m just curious how you’re dealing with the possibility of the EC2 instance failing.

Thanks,
Dan