Snafu User Documentation¶
Contents:
Purpose¶
Snafu aims to be the Swiss Army Knife of Serverless Computing due to its design principles. It should be easy to get going and use, full of features for both client-side and server-side Function-as-a-Service tasks, robust, flexible and most useful in combination with other tools and services.
Tutorial¶
In case you haven’t installed Snafu yet, you can do so directly from the Git sources, from the Python Package Index, or from the Docker Hub. Run one of the following commands on your command line (obviously without the leading $ sign). Note that Snafu runs perfectly fine from the Git repository without installation but all commands in this tutorial will need a preceeding dotslash (i.e. ./snafu instead of snafu).
$ git clone https://github.com/serviceprototypinglab/snafu.git
$ easy_install -Z snafu
$ docker run -ti jszhaw/snafu
Afterwards, try out Snafu on a source file containing a simple hello world function. Create one in Python as follows.
$ cat > /tmp/test.py << END
> def helloworld():return 'Hello World'
> END
And then, run Snafu on this file. It will automatically extract the function names and parameters from the file. There is only one function and therefore it is made available with the name test.helloworld which results from the name of the file and the name of the function. Furthermore, Snafu informs about its default configuration: All function invocations are logged into a CSV file (by default, snafu.csv in the execution directory). All functions are executed internally which requires them to be Python 3 functions. And the invocations are controlled by the interactive command-line interface (CLI) which asks for the function name and, in case there were any, the function parameters. The function result is printed after the timing information. Subsequently, the next function is asked for, at which point you can cancel the tool as a single hello world function is not terribly interesting after all.
$ snafu /tmp/test.py
» module: /tmp/test.py
function: test.helloworld
+ logger: csv
+ executor: inmemory
+ connector: cli
Function name:test.helloworld
[1492884560.138][139976878737152][function:test.helloworld]
[1492884560.138][139976878737152][result:Hello World]
[1492884560.138][139976878737152][time:0.002ms]
[1492884560.138][139976878737152][overalltime:0.008ms]
Hello World
Function name:^C
The following invocation syntax is useful for batch processing. Snafu is instructed to execute one specific function directly (-x) and to omit all unnecessary output in quiet mode (-q).
$ snafu -q -x test.helloworld /tmp/test.py
Now suppose you’ve got a number of functions stored at a major commercial cloud provider, such as IBM Bluemix OpenWhisk, Google Cloud Functions, or AWS Lambda. You can fill up your Snafu by importing the functions as follows. Note that the access to the provider needs to be configured beforehand, i.e. the respective command-line tools need to be working (test with either of these: aws lamdba list-functions, wsk list, gcloud beta functions list). In the example, functions are imported from Lambda which only supports Python 2, and are therefore (heuristically) upgraded to Python 3.
$ snafu-import --source lambda --convert
That went well, did it? The next step is then hosting these functions.
$ snafu-control
Now, you can redirect your command-line tools to Snafu and call all functions locally without incurring provider charges. Note that the Google Cloud CLI does not allow for such redirection and needs to be patched instead (see tools/patch-gcloud in the Git repository).
$ alias aws="aws --endpoint-url http://localhost:10000 --cli-read-timeout 0"
$ alias wsk="wsk --apihost http://localhost:10000"
Developer Information¶
Snafu consists not only of the presented CLI scripts, but primarily of Python modules within the Snafu package. Each module must be imported separately. For instance, to access the core Snafu functionality, run the following command:
>>> import snafulib.snafu
>>> dir(snafulib.snafu)
['Snafu', 'SnafuContext', 'SnafuFunctionSource', 'SnafuImport', 'SnafuRunner', '__builtins__', ...]
The most common use case is to instantiate a Snafu object and call its activate method which requires a list of files or directories to scan for functions and a calling convention (“any”).