This content is copyright of CloudCredo. © CloudCredo 2015. All rights reserved.
As a CF hero
I want my app to be resilient
So that random failures won't take it offline
Let's ship it
# From the training home directory:
$ cd 05-resilience/imperfect-app
$ cf push
...
urls: imperfect-app-votive-seeress.cfapps.io
...
You've shipped your new app!
The static website is handling the traffic wonderfully
Everyone wants to use your new app, but...
It crashes
Version 1 Sucks, But Ship It Anyway
Note: Coding Horror wisdom
Embrace failure & run many instances of the same app
$ cf scale imperfect-app -i 3
$ cf apps
name state instances memory disk urls
imperfect-app started 2/3 64M 256M imperfect..Note: Because CF is running your app, running many instances of it is just a command away
Notice the different IPs that your app instances are running on...
Even though each DEA runs on a different host, multiple app instances can end up on the same host
The more app instances you have, the less likely that they will be running on the same host
The next generation CF runtime - a.k.a. Diego - does a much better job in regards to evenly distributing the app instances
Many crashed instances,
app still available
$ cf app imperfect-app
state since cpu memory disk
#0 running 2015-11-02 0.0% 25.3M of 32M 66.9M of 128M
#1 down 2015-11-02 0.0% 0 of 0 0 of 0
#2 down 2015-11-02 0.0% 0 of 0 0 of 0$ watch cf apps # Watch app instances restart in real-time
Health Manager in DEA v2 (a.k.a. HM9K)
Health Check in Diego v3
Note: When an app instance crashes, the Health Manager dubbed HM9K will notice this and restart the app instance
My app needs more memory
$ cf events imperfect-app
description
index: 1, reason: CRASHED... Exited with status 255 (out of memory)$ cf scale imperfect-app -m 256M
My app needs more disk space
Do not use the app's disk for persistence
$ cf app imperfect-app
state since cpu memory disk
#0 running 2015-11-02 0.0% 24.4M of 256M 95.2M of 256M
#1 running 2015-11-02 0.0% 112.4M of 256M 224.2M of 256M
#0 running 2015-11-02 0.0% 24.3M of 256M 95.2M of 256M$ cf logs imperfect-app --recent
Errno::EDQUOT - Disk quota exceeded @ io_write - infinite-file:$ cf scale imperfect-app -k 1G
Note: Use a service for persistence, do not store any files in the container that runs the app
Scale instances, disk & memory
Combine multiple options in a single command
$ cf help scale
NAME:
scale - Change or view the instance count, disk space limit...
USAGE:
cf scale APP_NAME [-i INSTANCES] [-k DISK] [-m MEMORY] [-f]
OPTIONS:
-i Number of instances
-k Disk limit (e.g. 256M, 1024M, 1G)
-m Memory limit (e.g. 256M, 1024M, 1G)
-f Force restart of app without prompt
As a CF hero
I want my app to be resilient
So that random failures won't take it offline
Any questions?
Questions cannot be stupid. Answers can.
- Read the Twelve-Factor App
- Benchmark app with Load Impact (crash 1-2 instances)
- Read about Blue-Green deployment in CF
- Use cf-blue-green-deploy cli plugin
This content is copyright of CloudCredo. © CloudCredo 2015. All rights reserved.