Trying to understand fdbcli test

I’m very confused looking at this foundationdb/ at e2fa51103696d9f40cd554c82d894c694cef4ade · apple/foundationdb · GitHub

On line 211 it seems to be trying to get the pid of the fdbserver, but then it on line 215 it tries to match the pid against the port.

How the heck is that ever suppose to work? From what I can tell this test isn’t setting a port and its being randomly generated here foundationdb/ at e2fa51103696d9f40cd554c82d894c694cef4ade · apple/foundationdb · GitHub

Is it expected that there is some magic that forces fdbserver use a specific pid during the test?

If that is indeed the case, it doesn’t seem to be working on macos 11.6 x86 as these test are failing when i run ctest --output-on-failure -v:

The following tests FAILED:
         10 - single_process_fdbcli_tests (Failed)
         12 - single_process_external_client_fdbcli_tests (Failed)

I dumped the pid and port from the test and they do not match.

What am I missing here? Are the tests not functional on macos?

The code at line 211 is trying to get the process info of all fdbserver process.
We will get all fdbserver processes’ info from here.
Then line 215 we do filtering by using the port number to only get the fdbserver process info of the process in the cluster. So on line 215 we are not matching but filtering.

And for the context here, the test is for suspend command where the process is suspended for the specified time and then restart.
The test logic here is after running suspend we check the cluster does not have response but the pid of the process still exists. That’s the reason why we want to have the pid info here.

That doesn’t make sense. Here’s what the values I saw for pinfos and port

pinfos = ['5678']
port = '1234'
pinfo = list(filter(lambda x: port in x, pinfos))
print(pinfo) // prints []

as you can see pinfos has a pid of a single fdbserver and pinfo will always be empty in this case, failing the assert on line 216

As you said you are running on macOS,
I think it’s the difference of pgrep -a where on linux it will give the full command line with the port number in it.
I am not sure about its behavior on macOS which I guess like you posted it only gives the pid.
The test is used in the github CI where it’s running the docker environment so I never tested it on macOS.

I see. Using ubunutu pgrep does behave differently than on macos


$ pgrep -a node
9 node --optimize-for-size --gc-interval=100000 --always-compact --enable-source-maps /cocalc/nvm/versions/node/v14.17.4/lib/node_modules/@cocalc/project/bin/cocalc-project.js --hub-port 6000 --browser-port 6001 --kucalc --hostname=


$ pgrep -a node

Thanks for your help @sfc-gh-clin, I’ve fixed the issue in this PR fixed fdbcli suspend command test failing on macos by Pyrolistical · Pull Request #5729 · apple/foundationdb · GitHub