PyParis2017 / Tutorial transcript - Function-as-a-service : a pythonic perspective on severless computing, by Josef Spillner

Function-as-a-Service: A Pythonic Perspective
on Serverless Computing – Tutorial @ PyParis
2017
Josef Spillner
Zurich University of Applied Sciences, School of Engineering
Service Prototyping Lab (blog.zhaw.ch/icclab/)
8401 Winterthur, Switzerland
josef.spillner@zhaw.ch
June 9, 2017
Learning Objective
Learn on a practical level about different FaaS programming conventions,
runtimes, service providers and tools. Use some of the tools by yourself
to FaaSify simple Python applications.
1 Using Lambada
Snafu extracts your functions from code files.
Obtain Lambada directly from its source repository. You will furthermore
need Python 3.5 or more recent installed.
Listing 1: Obtaining Lambada
git clone https :// gitlab.com/josefspillner/lambada.git
Start by creating a file temperature.py. It contains one function and one
class which together form your «temperature application» (Listing 2).
Listing 2: Temperature application
def temperature ():
return 42.23
class CityTemperature :
def temperaturecity (self , city):
self.city = city
return 99
Now use Lambada to transform the contents into FaaS units which can still
be executed locally but are ready to be deployed in AWS Lambda (Listing 3).
1
transcript
type:
tutorial
distribution:
public
status:
final
initiative:
Service Tool-
ing

Listing 3: FaaSification process
./ lambada --local temperature.py
Then verify the generated Python file and notice its structure. Use any
editor of your choice on the newly generated file temperature_lambdafied.py.
Afterwards, check the functionality in the Python console. Run the following
(Listing 4).
Listing 4: FaaSified temperature application
python3
>>> import test_lambdafied
>>> dir( test_lambdafied )
>>> test_lambdafied .temperature ()
>>> test_lambdafied . temperature_remote (None , None)
>>> test_lambdafied . netproxy_handler ({"d": "{}", "name": "
temperaturecity ", "args": ["Winti"], "classname": "
CityTemperature "}, None)
Finally, it is time to deploy the code to AWS Lambda and execute it there.
You can run the command twice to verify that in the second case no redeploy-
ment happens (Listing 5). In parallel, you can verify the deployments in the
AWS Lambda Dashboard at https://<orgname>.signin.aws.amazon.com/console.
Listing 5: FaaSified temperature application
./ lambada temperature.py
./ lambada temperature.py # nothing happens , already deployed
2 Using Snafu
Snafu runs your functions in code files.
Start off with obtaining Snafu from its source repository.
Listing 6: Obtaining Snafu
git clone https :// github.com/ serviceprototypinglab /snafu
You can directly run Snafu without any configuration. It ships with a num-
ber of default functions in various languages. Notice how on first startup, the
interpreters for non-Python functions are compiled first. This means that while
the first startup is a bit slow, the second and subsequent ones are quote fast
(Listing 7).
Listing 7: Running Snafu
./ snafu
# type Ctrl+D
./ snafu # everything compiled
# type Ctrl+D again
2

Now run Snafu in C mode, meaning that the default interpreter is for func-
tions implemented in C. Select the sample C Fibonacci function and execute it
with a low parameter n (Listing 8).
Listing 8: Executing a C function
./ snafu -e c
# type function: fib_so.handler
# type argument: 10
# type Ctrl+D
Now remember your previous example functions used with Lambada. Obvi-
ously, Snafu can interpret this ﬁle as well. Execute Snafu on this ﬁle and explore
the execution (Listing 9).
Listing 9: Executing a Python function
./ snafu .../ path -to -lambada/temperature.py
For testing functions prior to deployment to AWS, it is convenient to use
Snafu’s control plane mode. In this mode, the default calling convention is the
same as in AWS Lambda (Listing 10).
Listing 10: Using the control plane
./snafu -control .../ path -to -lambada/temperature.py
# notice how no methods are selectable
# type Ctrl+D
# add to source: lambda_handler(event , context):return str(
temperature ())
# & run snafu -control command again
# open a second terminal , and run:
aws --endpoint -url http :// localhost :10000 lambda invoke --
function -name test. lambda_handler --payload ’{" event ":
""}’ /tmp/_out
cat /tmp/_out
3 Using Fission
Note: You will need to have Kubernetes installed with minikube + kubectl.
Ensure that the minikube context is used by kubectl. Furthermore, ensure that
your system user is in the group docker.
Listing 11: Fission script
# prepare
kubectl config use -context minikube
kubectl create -f http :// fission.io/fission.yaml
kubectl create -f http :// fission.io/fission -nodeport.yaml
export FISSION_URL=http ://$(minikube ip):31313
export FISSION_ROUTER=$(minikube ip):31314
3

# initiate
fission env create --name python --image fission/python -env
curl https :// raw. githubusercontent .com/fission/fission/
master/examples/python/hello.py > hello.py
# explore & deploy
cat hello.py
fission function create --name hello --env python --code
hello.py
# output: function ’hello ’ created
fission route create --method GET --url /hello --function
hello
# output: trigger ’35ad2fae -**** -**** -**** -679 d755fde8f ’
created
curl http :// $FISSION_ROUTER /hello
# user interface
kubectl create -f https :// raw. githubusercontent .com/fission/
fission -ui/master/docker/fission -ui.yaml
# output: deployment "fission -ui" created
# output: service "fission -ui" created
xdg -open http ://$(minikube ip):31319
4 Using OpenLambda (tentative)
Listing 12: OpenLambda script
git clone https :// github.com/open -lambda/open -lambda
cd open -lambda
sudo apt -get install libpython2 .7-dev
sudo make
bin/admin new -cluster pyparis
# returns lots of JSON
bin/admin workers -cluster pyparis
# output: Started worker: pid ...
bin/admin status -cluster pyparis
# output: Worker Pings: ...
cat quickstart/handlers/hello/lambda_func.py
# ... minimal Lambda -signature Python function
cp -r quickstart/handlers/hello pyparis/registry
curl -X POST localhost :8080/ runLambda/hello -d ’{" name ": "
Alice "}’
# -> permission denied !?
4

PyParis2017 / Tutorial transcript - Function-as-a-service : a pythonic perspective on severless computing, by Josef Spillner

Recommended

More Related Content

What's hot (19)

Similar to PyParis2017 / Tutorial transcript - Function-as-a-service : a pythonic perspective on severless computing, by Josef Spillner (20)

More from Pôle Systematic Paris-Region (20)

Recently uploaded (20)

PyParis2017 / Tutorial transcript - Function-as-a-service : a pythonic perspective on severless computing, by Josef Spillner