Practical Operation Automation with StackStorm

INTERNET MULTIFEED CO.Copyright ©
Practical Operation Automation with
StackStorm
Shu Sugimoto
Software Development Manager, JPNAP
2018-11-05(Mon)

What you will learn
• Why StackStorm is suitable for automating day to day
operation tasks
• The actual method that helps you implement automation
for your current procedures with StackStorm
• Will not cover
• Southbound implementation to network equipment
• All features of StackStorm
2

Background of “Automation”
• ”Automation” is becoming more and more important
• Business agility
• Time saving
• etc...
• In reality
• “We know that automation is important.”
• “We think now we put more effort into this ever.”
• “But its progress is far less than ideal.”
• Why?
3

Automation is difficult: Why?
• A: Your current operation is NOT computer friendly
• 1. Your procedures are so complicated that you can’t simply
write a shell script that does it
• Which also leads you having many partial scripts,
unmanaged, here and there
• 2. There exists steps that requires human interaction within
your procedure documents like:
• ”Check that the result is sane.”
• “Confirm the output is intended.”
• How can computer tell it’s “sane” or “intended”?
4

Automation is difficult: Why?
• A: Your current operation is NOT computer friendly
• -> “To achieve automation, we first need to rebuild our
whole operation from scratch...”
• => Scope become too huge, impossible to estimate, can’t
set proper goal, brain freeze
• StackStorm might help solving them
5

StackStorm aka st2
• Open source IFTTT-ish middleware/framework
• IF This Then That
6
It’s powerful even “Then That” part alone
https://www.slideshare.net/brocade/eventdriven-automation-devops-way-iot-73581697

How StackStorm fits in
• 1. Powerful Workflow engine
• It’s possible to implement a fairly complex procedure
7

st2 Workflow vs Shell script
8
Shell Script StackStorm Workflow
Image from tweet by StackStorm official Twitter account @Stack_Storm
https://twitter.com/stack_storm/status/684921149898113024

st2 Workflow vs Shell script
9
with-items: branch execution for all items in array
join: wait for all
loop
Super flexible, but easy to code

Workflow components
10
Workflow
Action

Workflow components
11
version: '2.0'
examples.mistral-branching:
description: >
A sample workflow that demonstrates how to use conditions
to determine which path in the workflow to take.
type: direct
input:
- which
tasks:
t1:
action: core.local
input:
cmd: "printf <% $.which %>"
publish:
path: <% task(t1).result.stdout %>
on-success:
- a: <% $.path = 'a' %>
- b: <% $.path = 'b' %>
- c: <% not $.path in list(a, b) %>
a:
action: core.local
input:
cmd: "echo 'Took path A.'"
publish:
stdout: <% task(a).result.stdout %>
b:
action: core.local
input:
cmd: "echo 'Took path B.'"
publish:
stdout: <% task(b).result.stdout %>
c:
action: core.local
input:
Workflow
Action
Action
Action

st2 Workflow
• Consists of Actions
• Defines a flow of your task by connecting Actions
• …in YAML
• Can take inputs (parameters)
• Consumed in workflow
• As an input to child action (mostly)
• Can return an output
• Returns result state
• Success/Failure
• Multiple engines supported
• Mistral v2
12

st2 Action
• Unit in workflow
• The place where actual work is done
• e.g. Creating directories, run `make`, etc
• Can take input/return output
• Returns result
• There are several ways to implement actions
• Write python code -> most popular
• Use built-in runners*
• Super useful built-in runner: `remote-shell-cmd`
13
* Actions are interpreted and run by corresponding runners
e.g. python action -> written in python, run by “python-script” runner

remote-shell-cmd runner
• `remote-shell-cmd`
• Built-in runner
• Takes following parameters as an input
• target hostname
• username
• ssh_key or password
• cwd
• cmd
• Runs cmd in cwd
• on target host as username
• by logging in with ssh
14

Example action backed by remote-shell-cmd
15
---
enabled: true
name: remote1
runner_type: remote-shell-cmd
parameters:
hosts:
default: 192.168.33.10
username:
default: vagrant
password:
default: vagrant
cwd:
default: /vagrant
cmd:
default: |
set -x
pwd
ls -al
df -h
root@9fe86b6dce75:/# st2 run demo.remote1
.
id: 5bdd72e9ecc69005aed541d4
status: succeeded
parameters: None
result:
192.168.33.10:
failed: false
return_code: 0
stderr: '+ pwd
+ ls -al
+ df -h'
stdout: '/vagrant
total 8
drwxr-xr-x 1 vagrant vagrant 128 Nov 3 02:13 .
drwxr-xr-x 23 root root 4096 Nov 1 15:53 ..
drwxr-xr-x 1 vagrant vagrant 128 Nov 2 23:58 .vagrant
-rw-r--r-- 1 vagrant vagrant 165 Nov 3 02:13 Vagrantfile
Filesystem Size Used Avail Use% Mounted on
udev 487M 0 487M 0% /dev
tmpfs 100M 4.4M 96M 5% /run
/dev/mapper/debian--9--vg-root 62G 1.3G 58G 3% /
tmpfs 499M 0 499M 0% /dev/shm
tmpfs 5.0M 0 5.0M 0% /run/lock
tmpfs 499M 0 499M 0% /sys/fs/cgroup
/dev/sda1 236M 37M 187M 17% /boot
vagrant 932G 111G 822G 12% /vagrant
tmpfs 100M 0 100M 0% /run/user/1000'
succeeded: true
remote1.yaml (defining custom action)

Example action backed by remote-shell-cmd
16
---
enabled: true
name: remote2
parameters:
hosts:
default: 192.168.33.10
username:
default: vagrant
password:
default: vagrant
cwd:
default: /
cmd:
default: |
set -eux
TMPDIR=$(mktemp -d)
cd $TMPDIR
git clone https://github.com/mtoyoda/sl
cd sl
make
sudo cp sl /usr/local/bin
# cleanup working directory
cd /
rm -Rf $TMPDIR
remote2.yaml
• Written in YAML
• Multiline command accepted
• Shell features accepted
• vars
• comments
• cmd substitution: $()
• etc
• password-less sudo accepted
• pseudo TTY allocation
If you want to run this action for
other host, you can simply do:
$ st2 run demo.remote2 hosts=192.0.2.1
hosts=192.0.2.1,192.0.2.2
It’s even possible to run on
multiple hosts simultaneously
just by:

st2 Workflow features
• Child action can be a workflow
• You can nest workflows in workflows
• No restriction in levels
• Action output can be chained to an input of subsequent
actions
17
A
W
A
A
W
A
A
A
1
2
3
4
5
6
78

Output/Input chaining
18
version: '2.0'
demo.input-output-chaining:
type: direct
tasks:
mktemp:
action: demo.remote-mktemp
publish:
tmpdir: "{{ jsonpath_query(task('mktemp').result, '*.stdout')[0] }}"
on-success:
- build
build:
action: demo.remote-build
input:
cwd: "{{ _.tmpdir }}"
on-success:
- cleanup
cleanup:
action: demo.remote-cleanup
input:
target_path: "{{ _.tmpdir }}"
---
enabled: true
name: remote-mktemp
parameters:
hosts:
default: 192.168.33.10
username:
default: vagrant
password:
default: vagrant
cmd:
default: mktemp -d
---
enabled: true
name: remote-build
parameters:
hosts:
default: 192.168.33.10
username:
default: vagrant
password:
default: vagrant
cmd:
default: |
git clone https://github.com/mtoyoda/sl
cd sl
make
sudo cp sl /usr/local/bin
input-output-chaining.yaml
remote-mktemp.yaml
remote-build.yaml

Other useful features
• Action execution concurrency policy
• You can enforces the number of executions that can run
simultaneously for a specified action
• Either delay/cancel
• Jinja templating in YAML
• Intended for parameter manipulation
• Datastore (st2kv)
• The place that you can store any key-value data
• Encryption support
• Config parameters, transient data that needs to be
shared between workflows
19

• It’s possible to implement a fairly complex procedure
• remote-shell-cmd helps converting existing steps in
procedure document into st2 actions
• Action can encapsulate a set of steps
• e.g.) git clone ~ make ~ make install
• Good isolation makes actions highly reusable
• There are many actions ready for use (Community
packs*)
• https://exchange.stackstorm.org/
• 100+ available
20

• 2. Inquiries feature
• Pause a workflow and wait for human interaction
• “Hey, does this look right?”
• “If so, please return true”
• “if not, please return false”
• Implemented as a built-in action “core.ask”
21

Inquiries
22
Pause here and wait for input
“Would you like to continue? (yes/no)”
Resume the workflow / abort
core.ask
abort!
yes no
Give a response

Inquiries
23
version: '2.0'
demo.inquiry-simple:
type: direct
tasks:
mktemp:
action: demo.remote-mktemp
publish:
tmpdir: "{{ jsonpath_query(task('mktemp').result, '*.stdout')[0] }}"
on-success:
- pause-workflow
pause-workflow:
action: core.ask
on-success:
- build
build:
action: demo.remote-build
input:
cwd: "{{ _.tmpdir }}"
on-success:
- cleanup
cleanup:
action: demo.remote-cleanup
input:
target_path: "{{ _.tmpdir }}"
root@9fe86b6dce75:/# st2 execution get 5bdf1631ecc6900824f95afd
id: 5bdf1631ecc6900824f95afd
action.ref: demo.inquiry-simple
parameters: None
status: paused
result_task: mktemp
result:
192.168.33.10:
failed: false
return_code: 0
stderr: ''
stdout: /tmp/tmp.bFbYga6wDz
succeeded: true
start_timestamp: Sun, 04 Nov 2018 15:54:25 UTC
end_timestamp:
+--------------------------+------------------------+----------------+
| id | status | task |
+--------------------------+------------------------+----------------+
| 5bdf1634ecc6900824f95b00 | succeeded (2s elapsed) | mktemp |
| 5bdf1636ecc6900824f95b02 | pending | pause-workflow |
+--------------------------+------------------------+----------------+
root@9fe86b6dce75:/# st2 inquiry respond 5bdf1636ecc6900824f95b02
continue (boolean): yes
Response accepted for inquiry 5bdf1636ecc6900824f95b02.

Inquiries
24
“What is your favorite editor?”
(vi/vim/emacs/nano)
core.ask
abort!
vi
You can even branch actions based on input value
Oops...
vim emacs nano

• 2. “Inquiries”
• With these features, you can start automating daily
operations without changing any existing processes or
tools
• StackStorm helps you “start small”
25

Our case
• Target: Changing configurations of monitoring servers
(ping/mrtg/etc...) when add/modify/delete-ing IXP
customer
26

300+ lines of diff to check
This example is rather easy
Excerpt of proc doc
300+ lines
“Is intended config added?”

Our case
• Target: Changing configurations of monitoring servers
(ping/mrtg/etc...) when add/modify/delete-ing IXP
customer
• Before
• There is a procedure document for human ops
• Steps summary
• ssh into specific server
• cd to tool dir
• Run `rake`
• Generate configs
• Check diff
• Run `rake deploy`
• Apply configs to servers
28

Workflow strategy
• Replace all steps with custom actions using remote-shell-
cmd runner
• Pause with core.ask when workflow reaches the point that
requires human decision
• Check diff
• (Plus) Send a diff to Slack
• So that operators can check it easily
• Straightforward 
29

New workflow
30
slack
core.ask
deploy
done
abort!
yes no
init
rake
---
name: "server_config_generator_rake"
runner_type: "remote-shell-cmd"
description: "Generate server-config with server-config-generator."
enabled: true
parameters:
scg_env:
type: string
immutable: true
default: "{{ st2kv.system.scg.config.scg_env }}"
env:
type: object
immutable: true
default:
SCG_ENV: "{{ scg_env }}"
cwd:
type: string
default: "{{ st2kv.system.scg.config.scg_directory | trim | d('/usr/local/mfeed/bin/server
cmd:
type: string
immutable: true
default: bash -lc "rake"
hosts:
type: string
immutable: true
default: "{{ st2kv.system.scg.config.scg_hostname }}"
username:
type: string
immutable: true
default: "{{ st2kv.system.scg.config.username | trim | d('mfeed', true) }}"
private_key:
type: string
immutable: true
default: "{{ st2kv.system.scg.config.ssh_key.remote_cmd }}"
sudo:
type: boolean
immutable: true
default: false

New workflow
31
Use `slack.files.upload` action from community
Diff is uploaded as snippet
slack
core.ask
deploy
done
abort!
yes no
init
rake

New workflow
32
“Does this diff look right? (yes/no)”
$ st2 inquiry respond 5bdbe0395c48de01de0f84cd -r
'{"continue": true}'
slack
core.ask
deploy
done
yes no
init
rake
abort!

New workflow
33
slack
core.ask
deploy
done
yes no
init
rake
---
name: "server_config_generator_deploy"
runner_type: "remote-shell-cmd"
description: "Deploy configs to servers"
enabled: true
parameters:
scg_env:
type: string
immutable: true
default: "{{ st2kv.system.scg.config.scg_env }}"
env:
type: object
immutable: true
default:
SCG_ENV: "{{ scg_env }}"
deploy_main:
type: boolean
default: false
description: "Choose a deploy target system. Can choose backup( = false ) or main( = true
cwd:
type: string
default: "{{ st2kv.system.scg.config.scg_directory | trim | d('/usr/local/mfeed/bin/server
cmd:
type: string
immutable: true
default: bash -lc "rake deploy_{% if deploy_main %}main{% else %}backup{% endif %}"
hosts:
type: string
immutable: true
default: "{{ st2kv.system.scg.config.scg_hostname }}"
username:
type: string
immutable: true
default: "{{ st2kv.system.scg.config.username | trim | d('mfeed', true) }}"
private_key:
type: string
immutable: true
default: "{{ st2kv.system.scg.config.ssh_key.remote_cmd }}"
sudo:
type: boolean
immutable: true
default: false
abort!

Findings
• We could implement our workflow in very short time
• Pretty straightforward thanks to `remote-shell-cmd`
and inquiries
• I’m confident that this approach is effective
• Everything is in YAML: Good
• We could apply the exact same methodology for
software development
• git
• Branch > PR > Code review > Merge
• CI/CD
• Staging/Production
• Disposable environment
• Easy to reproduce: just setup everything from git
• no “export/import”
34

Findings
• Development of st2 is active and open
• Fast release cycle: once in 3 months
• They widely accept PR from anyone
• You can find many active members at community Slack
• Direct channel to developers/product manager
• Many contributors who can help you
• Adopting StackStorm will not eliminate the need of
software engineers
• You still need them to achieve sustainable development
35

Conclusion
• With StackStorm, you can “small start” your long journey of
automation
• This can be achieved by its 1. powerful workflow engine,
and 2. inquiries feature
• Once you get there, it will naturally start advancing
• `core.ask` is where you should work on next
36

How to get started
• Building StackStorm environment into your dev machine
• vagrant-st2
• st2-docker
• (oneline installer)
• Tutorials
• Still does not exist a best one...
• https://github.com/StackStorm/st2-
docker/blob/master/docs/tutorial.md
• Official document
• https://docs.stackstorm.com
• For busy people: Skip to ”Actions”, “Workflows”, “Packs”
• Workflow examples
• https://github.com/stackstorm/st2/tree/master/contrib/examples
• Community Slack
• https://stackstorm.com/community-signup
37

StackStorm Tips
• You should use ”orquesta” workflow engine if you start now
• Although all examples in this presentation use mistral
• There are various reasons to this, but the major one is, orquesta is developed
by st2 team by own, mistral not (it’s a part of OpenStack project)
• Can expect much better support and faster bugfix
• Still in beta, but planned to be GA in Nov. 2018
• You should never include any sensitive data like passwords/private_keys in workflows
or actions
• Use st2kv or pack config to split them out
• You should avoid persisting any business data to st2kv
• Keep source of truth in other place
• Keep st2 disposable
• If you require HA deployment, you should check Kubernetes support
38

Practical Operation Automation with StackStorm

More Related Content

What's hot (20)

Similar to Practical Operation Automation with StackStorm (20)

Recently uploaded (20)

Practical Operation Automation with StackStorm