This is a Python script that will hit various API endpoints (in moderate volumes) using a provided PAT and report on success/failure rates. The tool will clean up after itself, deleting any repositories/issues/files/etc... created during testing.
This tool will specifically test functionality using newly created repositories. It is not intended to advise on the health of any existing repositories, and should not be used as the only source of truth for validating functionality of GitHub Enterprise Server.
- Python 3
requestsmodule ($ python -m pip install requests)- Supplied PAT must have
delete_repo, reposcopes
smoketest.py [-p] -pat PERSONAL_ACCESS_TOKEN [-t] -target GHES_URL [-n] -num NUM_REPOS [-debug]
Performs simple testing against GitHub Enterprise Server to ensure basic functionality. By default, each endpoint will be tested 10 times.
Required arguments:
-p, -pat The personal access token to use for access
-t, -target The URL of the target GitHub Enterprise Server instance (e.g. https://github.company.com)
Optional arguments:
-debug Display response output as JSON
-n, -num Number of repositories to create during testing (default 10)
Successful run:
$ python3 smoketest.py -p <pat> -t https://fakeserver.net -n 2
2021-11-16 18:58:48,253 INFO: Server at https://fakeserver.net/status appears to be up!
2021-11-16 18:58:48,444 INFO: Running as user1 - PAT auth confirmed working
2021-11-16 18:58:50,636 INFO: Loop smoketest_repo1 completed successfully.
2021-11-16 18:58:52,470 INFO: Loop smoketest_repo2 completed successfully.
2021-11-16 18:58:52,470 INFO: Testing completed successfully.
Failing run:
$ python3 smoketest.py -p <pat> -t https://fakeserver.net -n 2
2021-11-17 08:25:25,168 INFO: Server at https://fakeserver.net/status appears to be up!
2021-11-17 08:25:25,369 INFO: Running as user1 - PAT auth confirmed working
2021-11-17 08:25:25,755 ERROR: post request to user/repos failed! Expected a 201 response, but got a 422 response.
2021-11-17 08:25:26,423 ERROR: put request to repos/user1/smoketest_repo1/contents/testfile failed! Expected a 201 response, but got a 422 response.
2021-11-17 08:25:26,683 ERROR: Loop smoketest_repo1 completed with 2 errors.
2021-11-17 08:25:26,888 ERROR: post request to user/repos failed! Expected a 201 response, but got a 422 response.
2021-11-17 08:25:27,567 ERROR: put request to repos/user1/smoketest_repo2/contents/testfile failed! Expected a 201 response, but got a 422 response.
2021-11-17 08:25:27,842 ERROR: Loop smoketest_repo2 completed with 2 errors.
2021-11-17 08:25:27,842 ERROR: Testing completed with 4 errors - 50.0% failure rate. Please review logs!
Bad PAT:
$ python3 smoketest.py -p <bad_pat> -t https://fakeserver.net -n 2
2021-11-17 08:36:05,389 INFO: Server at https://fakeserver.net/status appears to be up!
2021-11-17 08:36:06,206 ERROR: PAT authentication failed with error: Bad credentials. Please check credentials.
We have rate limiting enabled. How many API calls does this make?
This script will make 42 API calls by default. 41 of these are subject to the rate limit, and 1 is not (the call to /status to ensure that the server is operational). Effectively, the number of calls will be equal to (<num value> * # of tests)+2.
It seems like some of this could be done more efficiently via GraphQL rather than the REST API. Why not use it?
The goal of this is not efficiency, but coverage. A single GraphQL call to do a significant amount of work only gives us a binary failure state. If we can spread that, instead, across tens/hundreds of API calls we'll get a better idea of how reliably GitHub Enterprise Server is able to respond. This is especially important for deployments that have a load balancer appliance sitting in front of GitHub Enterprise Server as we're more likely to hit many/all of the underlying servers in this test.
I'm getting spammed with
InsecureRequestWarningmessages when I run this. What's up?
You're probably using a self-signed certificate. I've already added verify=False to all of the API calls that are being done in this script to try to work around this in most cases, however you'll still get these annoying errors.
The best solution is to ensure you've got the CA for server added to your local system, however barring that you could run export PYTHONWARNINGS="ignore:Unverified HTTPS request" to disable this warning.
Obviously I'd prefer that you use a valid cert, but I know that's not always realistic.
Why Python and not Ruby?
I felt like working on a project in Python. 🤷
Can you add testing for X endpoint?
Sure - please feel free to fork and open a Pull Request! Fortunately, with requests in Python it's pretty trivial to add new functionality here, so I'd encourage it as a learning project for anyone interested. This should be as simple as adding a new api_call() invocation with the proper arguments.