Hello! I am creating a PoC for a Go monorepo, and I am trying out Pants. I got things working, and I have a Go package where I run some tests. This is a single file with 9 quite simple tests, so nothing big. The dependencies are quite big though. When I first ran the test, it took quite long, as analysing such big dependencies took a while. But, when I just rerun the exact same command without changing any code, it takes 22 seconds for it to run, and this happens every time. The command I run is

pants test libs/generators/azure:azure

, and it says

libs/generators/azure:azure succeeded in 0.91s (cached locally).

, which I interpret as the actual tests took less than a second, while the setup took 21 seconds, which seems a bit much. Am I doing something wrong, or is this expected? I attached a 30 second video just to show it in action.

So two things... Your pantsd is dying, which causes a lot of things to get invalidated unncessarily. Which pants version are you on? If you remove

.pants.d/pants.log

and then run it a few times, can you then share that log? Would be interesting to see why that happened. You might benefit from tuning the max memory, that is a common cause for spurious shutdowns. The logs will say if that is why, or another reason. Second, I'm quite sure you're seeing https://github.com/pantsbuild/pants/issues/20274... If you can do a two back-to-back runs with

pants -ldebug <...>

and share the logs that would help diagnose that.

Thank you for the reply! I think you are correct. I wasn't aware of the log, sorry. This is the

.pants.d/workdir/pants.log

after I deleted it and reran the command:

13:02:13.15 [INFO] pantsd 2.22.0 running with PID: 23200
13:02:13.18 [INFO] handling request: `test libs/generators/azure:azure`
13:02:22.19 [ERROR] The scheduler was invalidated: Exception('pantsd process 23200 was using 4189.11 MiB of memory (above the `--pantsd-max-memory-usage` limit of 4096.00 MiB).')
13:02:22.34 [ERROR] service failure for <pants.pantsd.service.scheduler_service.SchedulerService object at 0x10770b6a0>.
13:02:22.34 [INFO] Waiting for ongoing runs to complete before exiting...
13:02:33.64 [INFO] request completed: `test libs/generators/azure:azure`
13:02:33.64 [INFO] Server exiting with Ok(())
13:02:33.64 [INFO] Waiting for Sessions to complete before exiting...
13:02:33.64 [INFO] Waiting for shutdown of: ["scheduler_service_session", "store_gc_service_session", "pants_run_2024_10_02_13_02_14_324_d07f5bc721544172a468db293c35b6db"]
13:02:33.64 [INFO] Shutdown completed: "scheduler_service_session"
13:02:33.64 [INFO] Shutdown completed: "store_gc_service_session"
13:02:33.64 [INFO] Shutdown completed: "pants_run_2024_10_02_13_02_14_324_d07f5bc721544172a468db293c35b6db"
13:02:33.67 [INFO] Exiting pantsd
13:02:33.67 [WARN] File watcher exiting with: The watcher was shut down.

Turning the max memory to 8GiB instead of 4Gib solved the issue, and it now says

libs/generators/azure:azure succeeded in 0.91s (memoized).

instead.

Thank you! Glad to hear the memory bump seems like a solution. Re the other two logs, nothing stands out there... If you have the time, can you repeat the same experiment with

-ltrace

? I was hoping to see whether it was actually doing any work or just spend those 18 seconds on retrieving cached info. I have a sneaking suspicion most of those 4Gbs aren't actually used, they're just part of the dispatch for elidable work.

Of course! Added the two new runs as attachments.

Awesome, thank you! Will take a peek later, I assume these are even longer.

They are both 13 megabytes, so yeah, quite a lot 🙂

Ok, looked through it. For all I can tell, both of those runs are pulling from the cache, there's no actual work being done. It's just dispatching a metric ton of Pants internal analysis rules - ~15k.

Thank you for checking! Do you want me to add anything to the existing issue?

If the repo is open another repro is always helpful, but I think it's mostly a time issue at this point, someone has to do the (probably significant) refactor of the code here.

It is not open, unfortunately 😞 Anyways, thanks for helping! Tell me if there is anything more you want me to test.