Let's discuss the most common use cases of the Noop test framework and how it can be used.
In most cases you should setup the environment before using the Noop tests framework. The setup consists of three parts:
There is a wrapper script tests/noop/setup_and_diagnostics.sh that will try to do all these things. First, it will clone the fixtures repository unless it have already been cloned, then it will run tests/noop/noop_tests.sh with options -b and -B to create and use the bundle gems folder, -l options will enable Puppet modules update and -t option will initiate check procedures for paths and task library content.
If you are using RVM or are managing Ruby gems manually you are free to bypass this stem and clone fixtures repository manually.
Running tests/noop/noop_tests.sh without any options will try to execute all the spec tasks one by one. The list of the spec tasks will be generated by combining all the possible combinations of specs, hiera files and facts files that are allowed for this spec. Adding -p options allows you to review the list of tasks that will be run without actually running them.
Running tasks one by one will be a very time consuming process, so you should try to use multi-process run instead by providing the -j options with a number of processes that can be started simultaneously. In this mode you will not see the output of every RSpec process, but you will be able to get the combined result report of all test runs at the end of the process.
You can also monitor the progress of tasks by using the debug option -d. It will show you which tasks are starting and finishing and what shell commands and environment variables are used to run them.:
tests/noop/noop_tests.sh -j 24 -d
Will run all the spec task with debug output and keeping no more then 24 child processes all the time.:
Will output the list of tasks that is going to be run together with the facts and yaml files used.
There is also the run_all.sh shortcut script for this action.
In many cases you would want to run only a subset of tasks, up to only one task. You can use filters to do so. By providing the -s, -y and -f will allow you to set one or more specs, yams and facts that you want to use. The list of tasks will be build by filtering out everything else that you have not specified. Don't forget that you can use the -p option to review the list of tasks before actually running them.
List options -Y, -F, -S and -T can be used to view the list of all hiera yaml files, facts files, spec files and task files respectively. These lists are very helpful for finding out correct values for the filter options you want to use. Note, that using filter and list options together will allow you to see the filtered lists.:
Will list all available hiera yaml files.:
tests/noop/noop_tests.sh -y neut_vlan.compute.ssl.yaml -p
Will show you which task are going to run with this yaml file.:
tests/noop/noop_tests.sh -y neut_vlan.compute.ssl -s firewall/firewall -f ubuntu
Will run the firewall/firewall spec test with the provided yaml and facts file, but it will only work if this combination is allowed for this spec. Note, that you can either provide .yaml, _spec.rb, .pp extensions for yaml and spec files, or you can omit them and they will be found out on their own.:
tests/noop/noop_tests.sh -y neut_vlan.compute.ssl,neut_vlan.compute.nossl -s firewall/firewall,netconfig/netconfig -p
Filters can use used with a list of elements or can be given as a regular expression or a list of regular expressions:
./noop_tests.sh -p -s 'master/.*'
Will filter all tasks in the master group.:
./noop_tests.sh -p -s '^ceph.*,^heat.*,glance/db_spec'
Will filter all ceph, heat tasks and glance/db_spec task individually.:
./noop_tests.sh -p -s '^ceph.*' -y ceph
All ceph related tasks only on Hiera files which have Ceph enabled.
All globals files should already be precreated and commited to the fixtures repository and there is no need for you to create them again in the most cases. But, if you have made some changes to the existing yaml files or have added a new one, you should create the globals yamls again.
You can do it by running tests/noop/noop_tests.sh with -g option. It will set filters to run only the globals tasks as well as enabling the option to save the generated yaml files. Using -j option will make the process much faster.
There is also the run_globals.sh shortcut script for this action.
The framework builds the list of tasks to run by combining all allowed facts and yaml files for each spec file and creating a task for every combination.
By default, the list of yaml files, allowed for each spec will be determined by the intersection of node roles the spec should be run on (obtained from the tasks.yaml files used by the Nailgun to compile the deployment graph) and the hiera file roles (obtained from the hiera files themselves). The facts file will default to ubuntu value.
In most cases it would be better to manually specify the hiera files and, possibly, the facts files for your spec file, because running a task for every hiera file with the same roles would be redundant. On the other hand, if you know which Hiera files can cause a different behaviour of your task, and you want to test this behaviour in the different scenarios, you can explicitly specify the list of yaml files and facts files you need.
The list of Hiera files can be set by using the HIERA: commented annotation string followed by the list of hiera file names separated by the space character.:
# HIERA: neut_vlan.compute.ssl neut_vlan.compute.nossl
The list of facts files can be specified the same way using the FACTS: annotation.:
# FACTS: centos6 centos7
The list of task will contain this spec with all possible combinations of the specified Hiera and facts files. If you need to enter only the precise list of possible run combinations you can use the RUN: annotation.:
# RUN: (hiera1) (facts1) # RUN: (hiera2) (facts2)
It can be specified many times an all entered combinations will be added to the list.
There is also a way to use the reverse logic. You can specify the Hiera and facts yaml files that you want to exclude from the list instead of providing the list of included files.:
# SKIP_HIERA: neut_vlan.compute.ssl neut_vlan.compute.nossl # SKIP_FACTS: centos6
These yaml files will be excluded from the list of possible yaml files. If you have used both include and exclude options, the exclude option will have the priority over the include option. If there are no included Hiera files the list of Hiera files will be generated from the node roles.
The final annotation DISABLE_SPEC allows you to temporarily disable the spec from being seen the framework. It can use useful if you want to turn off a spec with run problems and fix them later without breaking the tests.:
The spec file with this annotation will be completely ignored.
In some cases you need a special set of facts values or the Hiera data for your task. If this values are very specific and are not useful for other tasks you can use override system instead of creating the new Hiera or facts yaml.
There are override folders inside the Hiera and facts folders. If you place a yaml file with the specific name to this folder, it will be used during the spec catalog compilation as the top level of Hiera's hierarchy. The values which are specified there will be used before the values in other yaml files. Hash values will be merged on the basic values and the matching key will be rewritten. Facts yamls work the same way by rewriting the basic values by the values specified in the override file.
Both yaml files should be named after the task name with path separator changed to the dash character. For example, the firewall/firewall task will use the override file name firewall-firewall.yaml and openstack-controller/keystone task will use the file name openstack-controller-keystone.yaml if these files are found in the override folders.
When the task manager runs the tasks they leave report files anf the manager can collect them to generate a combined test report seen at the end of the test process. These files can be found in the reports folder and re in json format.
You can use -r and -R options to load the saved reports from the previous run and display the report again, or to load reports and run the tasks that have previously failed after you have tried to somehow fix them.
You can use option -o to filter out only failed tasks and examples from the report and -O options to show only tasks without showing the individual examples. These options can be used together to show only failed tasks.
The task manager can also generate a test report in jUnit XML format using the -x options. It will be saves to the report.xml file in the reports folder of the fixtures repository. This file can be used by many tools to visualize the tests results, notably by the Jenkins CI.
There are several features that can be helpful during writing the initial spec for a task or when you are debugging a spec failure. Running tasks with -a options will show the report text about which files are being used in this task run and what files are found in Hiera and facts hierarchies.
Using -A option will output the entire compiled catalog in the Puppet DSL format. You can review its content and resource parameters to either find out what resources and classes are there or to see what values the parameters and properties actually have after all the catalog logic is processed. It's very helpful when you are debugging a strange task behaviour or writing a spec file.
The framework can also gather and report information about File resources that are being installed by Puppet. Using --save_file_resources options will dave the list of files that would be installed by the catalog and description about their source or template. Using --puppet_binary_files option will enable additional RSpec matcher that will fail if there are files and, especially, binary files being installed. These ones should be delivered by fuel packages.
Usually the spec files try to repeat the logic found in the tested manifests, receive the same set of resources and their parameters and compare them to the set of resources found in the compiled catalog. Then the matchers are used to check if the catalog contains what is expected from it to contain.
While this method works well in most cases it requires a lot of work and extensive expertise in the tasks' domain to write a correct and comprehensive set of spec for a task catalog. Specs also cannot detect if there are several new resources or properties that have not been described in the spec file.
Data-driven tests can offer an alternative way to ensure that there are no unwanted changes in the tasks catalogs. The idea behind them is building catalogs in human-readable format before and after the changes are made. Then these files can be compared and everything that have been changes will become visible.
Using the -V options will save the current catalog to the catalogs folder. These generated catalogs can be useful when reviewing complex patches with major changes to the modules or manifests. A diff for data changes may help a developer/reviewer examine the catalog contents and check that every resource or class are receiving the correct property values.
You can also use -v option to enable automatic catalog checks. It should be done after you have generated the initial versions and made some changes. Running the tests with this option enabled will generate the catalogs again and compare them to the saved version. If there are differences the test will be failed and you will be able to locate the failed tasks. Here is an example workflow one may use to examine a complex patch for data layer changes (we assume she is a cool ruby developer and use the rvm manager and bundler):
$ git clone https://github.com/openstack/fuel-library $ cd fuel-library $ rvm use ruby-2.1.3 $ PUPPET_GEM_VERSION=3.4.0 $ PUPPET_VERSION=3.4.0 $ ./tests/noop/setup_and_diagnostics.sh -B $ ./deployment/remove_modules.sh && ./deployment/update_modules.sh $ ./tests/noop/noop_tests.sh -V -j10 -b -s swift $ git review -d $swift_die_hard_patch_id $ ./tests/noop/noop_tests.sh -v -j10 -b -s swift
At this point, the reviewer will get the data diffs proposed to the swift modules and finish her thorough review. Note that the command ./tests/noop/setup_and_diagnostics.sh -B gets a clean state and sets things up, while removing and updating_modules is required to make things go smooth.
There are a number of environment variables used by either the task manager or by the specs themselves which can alter their behaviour and override the default or calculated values.
Many of this variables can be set by the Noop manager CLI options, or you can always export them externally.