Troubleshooting / common issues

Frequently asked questions (FAQs) and known issues you may encounter.

---

We love giving help!

See Getting Help. We would love to help!

If you are confused by something, likely someone else will run into the same issue. It is helpful for us to know what is going wrong so that we can improve Pants and improve this documentation.

Debug tip: enable stack traces and increase logging

Pants defaults to not displaying the full stack trace when it encounters an error. Pants also defaults to logging at the info level.

When you encounter an exception, it can help to use the global options --print-stacktrace and -ldebug, like this:

pants --print-stacktrace -ldebug <rest of your command>

Setting the option --pex-verbosity=9 can help debug exceptions that occur when building .pex files.

Once you have this stack trace, we recommend copying it into Pastebin or a GitHub Gist, then opening a GitHub issue or posting on Slack. Someone from the Pants team would be happy to help. See Getting Help.

Debug tip: inspect the sandbox with --keep-sandboxes

Pants runs most processes in a hermetic sandbox (temporary directory), which allows for safely caching and running multiple processes in parallel.

Use the option --keep-sandboxes=always for Pants to log the paths to these sandboxes, and to keep them around after the run. You can then inspect them to check if the files you are expecting are present.

pants --keep-sandboxes=always lint src/project/app.py
...
21:26:13.55 [INFO] preserving local process execution dir `"/private/var/folders/hm/qjjq4w3n0fsb07kp5bxbn8rw0000gn/T/process-executionQgIOjb"` for "Run isort on 1 file."
...

You can also pass --keep-sandboxes=on_failure, to preserve only the sandboxes of failing processes.

There is even a __run.sh script in the directory that will run the process using the same argv and environment variables that Pants would use.

Cache or pantsd invalidation issues

If you are using the latest stable version of Pants and still experience a cache invalidation issue: we are sorry for the trouble. We have not yet added a comprehensive goal to "clear all caches", because we are very interested in coming up with coherent solutions to potential issues (see for more information). If you experience a cache issue, please absolutely file a bug before proceeding to the following steps.

To start with, first try using --no-pantsd. If that doesn't work, you can also try --no-local-cache.

If --no-pantsd worked, you can restart pantsd, either by:

• Killing the pantsd process associated with your workspace. You can use ps aux | grep pants to find the PID, the kill -9 <pid>.
• Deleting the <build root>/.pids directory.

If this resolves the issue, please report that on the ticket and attach the recent content of the .pants.d/pantsd/pantsd.log file.

If restarting pantsd is not sufficient, you can also use --no-local-cache to ignore the persistent caches. If this resolves the issue, then it is possible that the contents of the cache (at ~/.cache/pants) will be useful for debugging the ticket that you filed: please try to preserve the cache contents until it can be resolved.

Pants cannot find a file in your project

Pants may complain that it cannot find a file or directory, even though the file does indeed exist.

This error generally happens because of the option pants_ignore in the [GLOBAL] scope, but you should also check for case-mismatches in filenames ("3rdparty" vs "3rdParty"). By default, Pants will read your top-level .gitignore file to populate pants_ignore, along with ignoring dist/ and any top-level files/directories starting with ..

To override something included in your .gitignore, add a new value to pants_ignore and prefix it with !, like the below. pants_ignore uses the same syntax as gitignore.

pants.toml

[GLOBAL]
pants_ignore.add = ["!folder/"]

Alternatively, you can stop populating pants_ignore from your .gitignore by setting pants_ignore_use_gitignore = false in the [GLOBAL] scope.

Import errors and missing dependencies

Because Pants runs processes in hermetic sandboxes (temporary directories), Pants must properly know about your dependencies to avoid import errors.

Usually, you do not need to tell Pants about your dependencies thanks to dependency inference, but sometimes dependency inference is not set up properly or cannot work.

To see what dependencies Pants knows about, run pants dependencies path/to/file.ext and pants dependencies --transitive.

Is the missing import from a third-party dependency? Common issues:

• Pants does know about your third-party requirements, e.g. missing python_requirements and go_mod target generators.
  • To see all third-party requirement targets Pants knows, run pants --filter-target-type=$tgt list ::, where Python: python_requirement, Go: go_third_party_package, and JVM: jvm_artifact.
  • Run pants tailor ::, or manually add the relevant targets.
• The dependency is missing from your third-party requirements list, e.g. go.mod or requirements.txt.
• The dependency exposes a module different than the default Pants uses, e.g. Python's ansicolors exposing colors.
  • Python: set the modules field and module_mapping fields.
  • JVM: set the packages field on jvm_artifact targets.
• Python: check for any undeclared transitive dependencies.

Is the missing import from first-party code? Common issues:

• The file does not exist.
  • Or, it's ignored by Pants. See the above guide "Pants cannot find a file in your project".
• The file is missing an owning target like python_sources, go_package, or resources.
  • Run pants list path/to/file.ext to see all owning targets.
  • Try running pants tailor ::. Warning: some target types like resources and files must be manually added.
• Source roots are not set up properly (Python and JVM only).
  • This allows converting file paths like src/py/project/app.py to the Python module project.app.
• Code generation such as Protobuf is not set up properly (Python and JVM only).
  • Generate missing targets so that produced modules could be found. If there are any Python files that are known to be created ad hoc only at runtime, you might consider using .pyi stub files for the modules to be discovered during dependency inference.

Common issues with both first and third-party imports:

• Ambiguity. >1 target exposes the same module/package.
  • If it's a third-party dependency, you should likely use multiple "resolves" (lockfiles). Each resolve should have no more than one of the same requirement. See Python and JVM.
  • If it's a first-party dependency, you may have unintentionally created multiple targets owning the same file. Run pants list path/to/file.ext to see all owners. This often happens from overlapping sources fields. If this was intentional, follow the instructions in the ambiguity warning to disambiguate via the dependencies field.
• Some target types like resources and files often need to be explicitly added to the dependencies field and cannot be inferred (yet).
• Multiple resolves (Python and JVM).
  • A target can only depend on targets that share the same "resolve" (lockfile).
  • Pants will warn when it detects that the import exists in another resolve. This usually implies you should either change the current target's resolve field, or use the parametrize() mechanism so that the code works with multiple resolves.
  • See Python and JVM.

When debugging dependency inference, it can help to explicitly add the problematic dependency to the dependencies field to see if it gets the code running. If so, you can then try to figure out why dependency inference is not working.

"Out of space" error: set an alternative tmpdir

It may be necessary to explicitly set the directory Pants uses as a temporary directory. For example, if the system default temporary directory is a small partition, you may exhaust that temp space.

Use the global option local_execution_root_dir to change the tmpdir used by Pants.

pants.toml

[GLOBAL]
local_execution_root_dir = "/mnt/large-partition/tmpdir"

"No space left on device" error while watching files

On Linux, Pants uses inotify to watch all files and directories related to any particular build. Some systems have limits configured for the maximum number of files watched. To adjust the limit on file watches, you can run:

echo fs.inotify.max_user_watches=524288 | sudo tee -a /etc/sysctl.conf && sudo sysctl -p

How to change your cache directory

You may change any of these options in the [GLOBAL] section of your pants.toml:

Option	What it does	Default
local_store_dir	Stores the results of running subprocesses and of some file operations.	~/.cache/pants/lmdb_store
named_caches_dir	Stores the caches for certain tools used by Pants, like PEX's cache for resolving Python requirements.	~/.cache/pants/named_caches
pants_workdir	Stores some project-specific logs; used as a temporary directory when running pants repl and pants run. This is not used for caching. This must be relative to the build root.	<build_root>/.pants.d/
pants_distdir	Where Pants writes artifacts to, such as the result of pants package. This is not used for caching; you can delete this folder and still leverage the cache from local_store_dir. This must be relative to the build root.	<build_root>/dist/

For local_store_dir and named_caches_dir, you may either specify an absolute path or a relative path, which will be relative to the build root. You may use the special string %(homedir)s to get the value of ~, e.g. local_store_dir = "%(homedir)s/.custom_cache/pants/lmdb_store".

It is safe to delete these folders to free up space.

You can also change the cache used by the pants script described in Installing Pants, which defaults to ~/.pants/cache/setup. Either set the environment variable PANTS_SETUP_CACHE or change the Bash script directly where it defines PANTS_SETUP_CACHE. You may use an absolute path or a path relative to the build root.

BadZipFile error when processing Python wheels

This can happen if your temporary directory (/tmp/ by default) is not on the same filesystem as ~/.cache/pants/named_caches, and is caused by the fact that pip is not concurrency-safe when moving files across filesystems.

The solution is to move ~/.cache/pants, or at least the named_caches_dir(see above), to the same filesystem as the temporary directory, or vice versa.

Issues packaging AWS CDK into a PEX

If you get errors like ModuleNotFoundError: No module named 'aws_cdk.asset_awscli_v1, set execution_mode="venv" and venv_site_packages_copies=True on your pex_binary target.

This ensures that the aws_cdk subpackages are properly nested under the parent package, despite those distributions not being configured as namespace packages.

"Double requirement given" error when resolving Python requirements

This is an error from pip, and it means that the same 3rd-party Python requirement—with different version constraints—appears in your dependencies.

You can use pants peek to help identify why the same requirement is being used more than once:

Shell

# Check the `requirements` key to see if it has the problematic requirement.
pants --filter-target-type=python_requirement peek ::

macOS users: issues with system Python interpreters

The macOS system Python interpreters are broken in several ways, such as sometimes resulting in:

ERROR: Could not install packages due to an EnvironmentError: [Errno 13] Permission denied: '/Library/Python/3.7'

You can set the option interpreter_search_paths in the [python] scope to teach Pants to ignore the interpreters in /usr/bin. See here for more information.

"Too many open files" error

You may encounter this error when running Pants:

pants count-loc helloworld/greet/f.py

ERROR: Could not initialize store for process cache: "Error making env for store at \"/Users/pantsbuild/.cache/pants/lmdb_store/processes/2\": Too many open files"

(Use --print-exception-stacktrace to see more error details.)

This sometimes happens because Pants uses lots of file handles to read and write to its cache at ~/.cache/pants/lmdb_store; often, this is more than your system's default.

This can be fixed by setting ulimit -n 10000. (10,000 should work in all cases, but feel free to lower or increase this number as desired.)

Tip: permanently configuring ulimit -n

We recommend permanently setting this by either:

1. Adding ulimit -n 10000 to your pants.bootstrap script.
2. Adding ulimit -n 10000 to your global .bashrc or equivalent.

The first two approaches have the benefit that they will be checked into version control, so every developer at your organization can use the same setting.

macOS users: avoid ulimit unlimited

Contrary to the name, this will not fix the issue. You must use ulimit -n instead.

Controlling (test) parallelism

When adopting Pants for your tests you may find that they have issues with being run in parallel, particularly if they are integration tests and use a shared resource such as a database.

To temporarily run a single test at a time (albeit with reduced performance), you can reduce the parallelism globally:

pants --process-execution-local-parallelism=1 test ::

A more sustainable solution for shared resources is to use the [pytest].execution_slot_var option, which sets an environment variable which test runs can consume to determine which copy of a resource to consume.

Snap-based Docker

In recent Ubuntu distributions, the Docker service is often installed using Snap. It works mostly same as a normal installation, but has an important difference: it cannot access the /tmp directory of the host because it is virtualized when Snap starts the Docker service.

This may cause problems if your code or tests ry to create a container with a bind-mount of a directory or file under the current working directory. Container creation will fail with "invalid mount config for type "bind": bind source path does not exist", because Pants' default local_execution_root_dir option is /tmp, which the Snap-based Docker service cannot access.

You can work around this issue by explicitly setting [GLOBAL].local_execution_root_dir to a directory outside the system /tmp directory, such as "%(buildroot)s/tmp".

Using pants on self-hosted GitHub actions runner

Setting up pants to run with Python executables provided by setup-python will not work on vanilla actions runner setup. This is due to the known limitation of pants which does not allow leaking arbitrary environment variable (in this case LD_LIBRARY_PATH for us) when evaluating dependency inference rules. If you fall into this situation, you will face an error complaining about missing shared object files, like this:

/home/ubuntu/.cache/python-tools/Python/3.11.3/x64/bin/python3.11: error while loading shared libraries: libpython3.11.so.1.0: cannot open shared object file: No such file or directory

One of the workaround to fix this issue is setting up python tool cache files at /opt/hostedtoolcache directory. This is the default path which setup-python action uses to download relevant files on hosted GitHub actions runner. Overriding tool cache download directory can be achieved by following setup-python documentation.