I launched my first "production" EC2 instance almost three years ago, on July 15, 2008. For my purposes, production includes hosting my personal blog and writing code for my AWS book, as well as a host for random development projects that I putter around with from time to time.
I am happy to report that my instance reached 1000 days of uptime over the weekend:
One of these days I'll upgrade to a more modern instance (this one predates EBS) but I'm still quite happy with this one and I'll keep it running as long as possible.
EC2 has certainly come a long way in just 1000 days. Here are some of the highlights:
- August 2008 - Elastic Block Storage (EBS).
- October 2008 - EC2 production status, SLA, Windows.
- December 2008 - EC2 in Europe.
- January 2009 - AWS Management Console.
- March 2009 - EC2 Reserved Instances.
- May 2009 - Elastic Load Balancing, Auto Scaling, and CloudWatch.
- August 2009 - Virtual Private Cloud.
- October 2009 - High Memory Instances.
- December 2009 - EC2 in Northern California, Boot from EBS, and Spot Instances.
- February 2010 - High Memory Extra Large Instances.
- April 2010 - EC2 in Singapore.
- May 2010 - Virtual Private Cloud in Europe.
- July 2010 - Cluster Compute Instances.
- September 2010 - Micro Instances, Filtering, Resource Tagging, Idempotent Instance Creation, Amazon Linux AMIs.
- October 2010 - SUSE Linux Enterprise Server.
- November 2010 - AWS Free Usage Tier.
- January 2011 - AWS Elastic Beanstalk.
- February 2011 - AWS CloudFormation.
- March 2011 - AWS in Tokyo, EC2 Networking, Dedicated Instances.
Jeff;
OM that is long.
Just reading your book now.
Thanks for your sharing.
Posted by: Lambert | April 11, 2011 at 11:14 AM
i got one at 923 days running loads simple website ,tomcat serving it up. I am continually amazed by AWS. I remember reading those comments by people saying dont trust the "cloud" for reliability etc etc 923 days later I cant complain !
Posted by: James McArthur | April 11, 2011 at 12:29 PM
Hmmm. I thought the normal hardware replacement cycle was three years, Jeff. Does that mean that your instance will finally be restarted after 1,095 days?
Posted by: Geoffarnold | April 11, 2011 at 01:29 PM
That will certainly comfort those who, like me, spend hours each week restarting instances which:
- crash mysteriously (won't even ping anymore),
- lose EBS volumes (can still SSH on the instance, "vmstat 1" shows 100% wait state and 0 iops on one of the EBS),
- refuse to shutdown (the instance remains in "stopping" state for a few hours),
- etc.
(And I only mentioned the failures which are certainly not to blame on our applications!)
Now, since this seems to be a lonely instance, it would be interesting to know how you implement the persistence of the data: everything into RDS and S3? Regular backups? Or you just happen to pray every morning for its incredible uptime to go on raising? :-)
Posted by: Jérôme Petazzoni | April 12, 2011 at 08:27 AM
I have the same question. Does that mean that they have not been able to update the O/S on the host server for the last three years?
Posted by: Brian Leffler | April 12, 2011 at 09:10 AM
Just out of curiosity, what has it cost you for those 1000 days?
Posted by: Lucas | April 14, 2011 at 08:16 AM
Brian - Xen (which I believe is what EC2 uses) allows vms to be migrated between hosts pretty much on the fly, precisely so that host machines can be upgraded, replaced etc. without disturbing the vms they host.
Posted by: Fred | April 14, 2011 at 09:27 AM
I got an instance up since 673 days
Posted by: sajal | April 15, 2011 at 11:39 AM
1000 days on ephemeral storage - that's what I call steel nerves :)
Posted by: Robert | April 20, 2011 at 03:16 AM
How is your instance doing today? Did you jinx everyone in Virginia with this post?
Posted by: Wes Horner | April 21, 2011 at 11:21 AM