Making Your ColdFusion Apps Highly Available

What happens when you move your web application from a single server to a cluster of servers? What do you need to think about in order to achieve 99.99% (or higher) uptime? This presentation, given at NCDevCon 2012, discusses those issues and gives an overview of Amazon EC2 as a possible deployment solution for your high-availability web app cluster.

Transcript

1. Making Your ColdFusion Apps Highly Available Brian Klaas Johns Hopkins Bloomberg School of Public Health bklaas@jhsph.edu @brian_klaas
3. How much downtime can you afford?
4. 99% = 14 minutes/day
5. 99.9% = 1.4 minutes/day
6. 99.99% = 8.6 seconds/day
7. 99.999% = 1 second/day
8. $
9. How much availability can you afford?
10. Um, hey, I’m not Google.
11. Scale out
12. Replication + Redundancy = Availability
13. Virtualization is your friend.
14. Network Changes Servers Database Code
15. Changes to App Code
16. != x
17. <cffile action = "upload" fileField = "profileImage" destination = "c:\files\upload\" /> X
18. Use network shares or Amazon S3.
19. CF + IIS CF + IIS CF + IIS Load Balancer File Share Database Static Asset Domain
20. <cfset pathToFiles = expandPath(" /fileShareDirectory/uploads" /> <cffile action = "upload" fileField = "profileImage" destination = "#pathToFiles#" />
21. 1. Set up mapped directory to the le share at the OS level. 2. Make sure CF has permission to access the share. Two Tips
22. What server am I on?
23. <cfset var inetAddressObj = createObject("java", "java.net.InetAddress") /> <cfset globalVars.machineName = inetAddressObj.localhost.getCanonicalHostName() /> Add to App Startup Code
24. Environment Con g ‘Cause you’re really, really, really going to need a testing environment.
25. ▪ Development, testing, production? ▪ Compare server name Environment Con g
26. ▪ Sticky sessions ▪ Database persistence ▪ Out-of-process or distributed cache Handling Sessions
27. Basic Monitoring
29. (Very) Basic Monitoring 1. Loop through list of servers to check 2. Make <cfhttp> call to a “heartbeat” event on each 3. Result = error, send an email/SMS alert
30. Changes to Database Setup
31. Replicate
32. CF + IIS CF + IIS CF + IIS Load Balancer File Share Static Asset Domain Database Master Database Slave
33. Distributed Databases eep
34. Changes to Server Con g
35. CF + IIS CF + IIS CF + IIS Load Balancer File Share Database Static Asset Domain
36. CF + IIS CF + IIS CF + IIS Load Balancer File Share Static Asset Domain Database Master Database Slave
37. CF + IIS CF + IIS CF + IIS Load Balancer Static Asset Domain Database Master Database Slave File Share 1 File Share 2
38. CF + IIS CF + IIS CF + IIS Load Balancer Database Master Database Slave File Share 1 File Share 2 Static Asset 1 Static Asset 2
39. CF + IIS CF + IIS CF + IIS Load Balancer Database Master Database Slave File Share 1 File Share 2 Scheduled Task Server Static Asset 1 Static Asset 2
40. Dev
41. CF + IIS File Share Database Static Asset Domain Dev
42. yymkw@Flickr What’s your patch strategy?
43. How do I get code there? FTP Ant Chef Puppet Jenkins
44. Changes to the Network
45. CF + IIS CF + IIS CF + IIS Load Balancer Database Master Database Slave File Share 1 File Share 2 Scheduled Task Server Load Balancer Static Asset 1 Static Asset 2
46. And what if the metal blows up?
47. Node 1 Node 2 Load Balancer Load Balancer CF + IIS CF + IIS CF + IIS Database Master Database Slave File Share 2 File Share 1 Static Asset Domain
48. And what if the data center blows up?
49. Considering AWS* * Applies to other cloud service providers too.
50. AWS* = unlimited virtualized servers in data centers around the globe. * And friends.
51. EC2
52. AMI on EC2
53. EBS or S3 Code Assets
54. Amazon RDS Databases: Your own MySQL Oracle MS SQL Server NoSQL DB MongoDB Cassandra CouchDB Riak Redis AZ Dynamo DB AZ Simple DB EC2 + EBS
55. ELB
56. EC2 + AMI EC2 + AMI EC2 + AMI ELB DB Master EC2 + AMI DB Slave EC2 + AMI Task EC2 + AMI EBS EBS EBS User Content S3 Static Asset S3 EBS EBS EBS EBS EBS
57. IOPS
58. Cache, cache, cache.
59. Everything fails.
60. You are responsible for redundancy.
61. Autoscale. CloudWatch
62. Elastic Beanstalk Or another PaaS provider.
63. Use multiple availability zones. Region = data center around the globe Availability zone = segment of a data center
64. Uh, so is it worth it?
65. Meta–Network Availability
68. How much availability can you afford?
69. Thank you! Brian Klaas Johns Hopkins Bloomberg School of Public Health bklaas@jhsph.edu @brian_klaas www.iterateme.com
71. Resources ▪ Amazon AWS aws.amazon.com ▪ Amazon Elastic Compute Cloud aws.amazon.com/ec2/ ▪ Amazon Machine Images aws.amazon.com/amis/ ▪ Amazon Elastic Block Storage aws.amazon.com/ebs/ ▪ Amazon Simple Storage Service (S3) aws.amazon.com/s3/
72. Resources ▪ Amazon Relational Database Service aws.amazon.com/rds/ ▪ Amazon Elastic Load Balancing aws.amazon.com/elasticloadbalancing/ ▪ Amazon Elastic Beanstalk (Autoscaling) aws.amazon.com/elasticbeanstalk/ ▪ Amazon ElastiCache aws.amazon.com/elasticache/ ▪ Amazon DynamoDB (NoSQL Service) aws.amazon.com/dynamodb/
73. Resources ▪ Amazon Route 53 (DNS Service) aws.amazon.com/route53/ ▪ Amazon Autoscale aws.amazon.com/autoscale/ ▪ The Official Word on CF10 Licensing Changes blogs.coldfusion.com/post.cfm/coldfusion-10-eula ▪ Jelastic (PaaS Vendor that Supports ColdFusion) jelastic.com ▪ Net ix’s Chaos Monkey github.com/Net ix/SimianArmy
74. Resources ▪ Apache ANT ant.apache.org ▪ Puppet puppetlabs.com ▪ Chef www.opscode.com/chef/ ▪ FusionReactor www.fusion-reactor.com ▪ Database Sharding www.codefutures.com/database-sharding/