Troubleshooting#

If you encournter an issue that is not listed here, you can obtain free support for Anaconda through the Anaconda community. For Anaconda installation or technical support options, visit our support offerings page.

Having problems with Anaconda Navigator? Find help in the Navigator troubleshooting guide.

General issues#

403 error#

Cause

A 403 error is a generic Forbidden error issued by a web server in the event the client is forbidden from accessing a resource.

The 403 error you are receiving may look like the following:

Collecting package metadata (current_repodata.json): failed

UnavailableInvalidChannel: The channel is not accessible or is invalid.
  channel name: pkgs/main
  channel url: https://repo.anaconda.com/pkgs/main
  error code: 403

You will need to adjust your conda configuration to proceed.
Use `conda config --show channels` to view your configuration's current state,
and use `conda config --show-sources` to view config file locations.

There are several reasons a 403 error could be received:

The user has misconfigured their channels in their configuration (most common)
A firewall or other security device or system is preventing user access (second most common)
We are blocking their access because of a potential terms of service violation (third most common)

Solution

Reset your default channel configuration by running the following command:

conda config --remove-key default_channels

When conda is first installed, the default channels it uses to install packages are https://repo.anaconda.com/main and https://repo.anaconda.com/r (as well as https://repo.anaconda.com/msys2 for Windows operating systems).

If your other channels require a token, install or upgrade the conda-token tool by running the following command:

# The --freeze-installed flag ensures that no already-installed package dependencies are updated
conda install --freeze-installed conda-token

Then, re-apply the token and configuration settings:

# Replace <TOKEN> with your token
conda token set <TOKEN>

If this doesn’t resolve the issue, Anaconda recommends consulting our Terms of Service error page.

HTTP 000 CONNECTION FAILED#

If you receive this error message, run the following command:

conda config --set ssl_verify false

Using Anaconda behind a firewall or proxy#

Corporate security policies may prevent a new Anaconda installation from downloading packages and other functionality that requires connecting to an external server. To make external connections you may need to connect to a firewall/proxy. Additionally, your IT team may need to allow connections to https://anaconda.org and https://repo.anaconda.com as these are the main package repositories.

Solution

To add the proxy information you will need to add two entries to your .condarc file located in the user’s home directory. This information should be made available by your IT team and may contain a username and password that is included in the URL. Read more about the .condarc configuration.

Example configuration

channels:
- defaults

proxy_servers:
  http: http://username:[email protected]:8080
  https: https://username:[email protected]:8443

In some situations it may be necessary to export the HTTP_PROXY and HTTPS_PROXY environment variables.

Windows

set HTTP_PROXY=http://username:[email protected]:8080
set HTTPS_PROXY=https://username:[email protected]:8443

MacOS/Linux

export HTTP_PROXY=http://username:[email protected]:8080
export HTTPS_PROXY=https://username:[email protected]:8443

If these steps have not allowed connections you should speak to your IT team to verify that security policies are not blocking connections to https://anaconda.com and https://repo.continuum.io.

InsecurePlatformWarning error#

Cause

“InsecurePlatformWarning” appears only when the installed version of Python is older than version 2.7.9. This message warns only that the validity of the SSL connection is not being verified. It should not affect your package downloads.

Solution

To resolve this, install the updated package ndg-httpsclient:

conda install ndg-httpsclient

Note

When initially installing this package, you receive the SSL warning again. Once it is installed, the package will prevent the warnings.

Anaconda search error: not recognized as an internal or external command/unrecognized arguments#

Cause

If anaconda-client is not installed and you search for a package on anaconda.org using the Anaconda search command:

# Replace <PACKAGENAME> with the name of a package you want to search for
anaconda search -t conda <PACKAGENAME>

You will receive the following error message:

Windows

'anaconda' is not recognized as an internal or external command,
operable program or batch file.

macOS/Linux

usage: conda [-h] [-V] command ...
conda: error: unrecognized arguments: -t <PACKAGENAME>

Solution

To resolve the error, install anaconda-client:

conda install anaconda-client

Then search for a package:

anaconda search -t conda <PACKAGENAME>

Conda: command not found on macOS or Linux#

Cause

The conda shell function is not available or is not working properly. Some causes:

You haven’t started a new shell after installing Anaconda/Miniconda (assuming you allow it to modify your startup script).
You didn’t allow the installer to modify your startup script.
You have set auto_activate_base to false. You need to run conda activate <ENV>, replacing <ENV> with the environment you want active. If you do not specify an environment the default is base.
Conda has been corrupted, usually by a change in the Python package (e.g. 3.6->3.7).

Solutions

If your install is new, close and reopen your terminal application. If you want your terminal to remain open, you can source your ~/.bash_profile or ~/.bashrc file. When you “source” a script, your terminal re-reads it and applies the changes, which usually only happens when the terminal is first opened.:

# Replace bash_profile with bashrc on Linux and Windows Subsystem for Linux
. ~/.bash_profile

—

To initialize conda in your current terminal, run the following command:

<PATH-TO-CONDA-INSTALL>/bin/conda init

For instance, if you installed the Anaconda distribution to the default location on macOS or Linux, your command would look like the following:

~/anaconda3/bin/conda init

—

To see the value for auto_activate_base, run the following command:

conda config --describe auto_activate_base

If your terminal returns true, this means that conda is not automatically activating your base environment when you start a new shell. This behavior emulates your system Python, and some users prefer to have their conda environment be inactive until they need it. However, this is not conda’s default behavior after installation.

To change the value of auto_activate_base, run the following command:

# Replace <TRUE-OR-FALSE> with true or false
conda config --set auto_activate_base <TRUE-OR-FALSE>

If you have auto_activate_base set as false, the conda command will still be available as a shell function, but your base environment will not be active when a new shell is started. To activate your base environment, run conda activate.

Conda: Channel is unavailable/missing or package itself is missing#

Cause

After a user has configured their .condarc for either Professional or Business, in some cases they are unable to install packages. They may receive an error message that the channel or package is unavailable or missing.

Solution

One potential fix for all of these is to run the following command:

conda clean -i

This will clear the “index cache” and force conda to sync metadata from the repo server.

Collecting package metadata (repodata.json): - Killed#

Cause

When installing or searching for a package, you may see the process end abruptly with a “Killed” message:

$ conda install numpy
Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Collecting package metadata (repodata.json): - Killed

This may be because your system lacks the sufficient disk space or memory to complete the process.

Solution

Verify that you have enough disk space and memory on your system to install and use Anaconda packages. The minimum system requirements for Miniconda and Anaconda installers can be found in the conda user guide.

Linking problems when Python extensions are compiled with gcc#

Cause

When compiling Python extensions with gcc on Windows, linking problems may result.

Solution

To resolve these linking problems, use the conda package libpython, a mingw import library that Anaconda builds and includes with the Anaconda Distribution.

Error message: Unable to remove files#

When trying to update or install packages with conda, you may see an error message such as:

Error: Unable to remove files for package: <package-name>
Please close all processes running code from conda and try again.

Cause

This may be caused by a file lock issue.

Solution

Before updating or installing any packages with conda, be sure to terminate any running Anaconda processes, such as Navigator, Spyder, or IPython.

You can also force the installation of the package:

# Replace <PACKAGENAME> with the name of the package
# you want to install
conda install -f <PACKAGENAME>

Using 32- and 64-bit libraries and CONDA_FORCE_32BIT#

To work with both 32- and 64-bit libraries, Anaconda recommends that you have two separate installs: Anaconda32 and Anaconda64 or Miniconda32 and Miniconda64.

When working with both versions, add the path to your installer files to the PATH.

Caution

Always specify which version you want to work with because mixing 32- and 64-bit packages can cause problems in your environment.

To get the information about conda, including your PATH, run the following command:

conda info -a

Using force_32bit is not recommended because it forces 32-bit packages to be installed in the environment, but does not force 32-bit libraries to load at runtime.

force_32bit should be used only when running conda-build to build 32-bit packages on a 64-bit system.

Windows-specific issues#

Cannot see Anaconda menu shortcuts after installation on Windows#

After installing on Windows, the Anaconda Prompt and Anaconda Navigator shortcuts do not appear in the Windows Start menu.

Cause

This may be caused by the way Windows updates the Start menu, or by having multiple versions of Python installed that are interfering with one another. Existing Python installations, installations of Python modules in global locations, or libraries that have the same names as Anaconda libraries can all prevent Anaconda from working properly.

Solution

If Start menu shortcuts are missing, Microsoft recommends rebooting your computer or restarting Windows Explorer.

If that doesn’t work, clear your $PYTHONPATH environment variable, if applicable, and re-install Anaconda.

Other potential solutions are covered in the “Conflicts with system state” section of the following blog post.

Windows error: Failed to create Anaconda menus or Failed to add Anaconda to the system PATH#

During installation on a Windows system, a dialog box appears that says “Failed to create Anaconda menus, Abort Retry Ignore” or “Failed to add Anaconda to the system PATH.” There are many possible Windows causes for this.

Solution

Try these solutions, in order:

Do not install on a PATH longer than 1024 characters.
Turn off anti-virus programs before installing, then turn the anti-virus programs back on.
Uninstall all previous Python installations.
Clear all PATHs related to Python in sysdm.cpl file.
Delete any previously set up Java PATHs.
If JDK is installed, uninstall it.

Note

To find your path variable:

Search for “environment variables” in the search in your taskbar.
Select Edit the System Environment Variables in the sidebar.
Click Environment Variables in the System Properties dialog.
Highlight the path variable for your user.
Click Edit.

Anaconda interfering with other software on Windows#

Cause

If a user chooses to add Anaconda to the Windows PATH, this can cause programs to use the new Anaconda versions of software (like Python) and not the versions that were already in place. In some cases, this can cause incompatibility and errors.

Solution

Anaconda recommends against adding Anaconda to the Windows PATH manually. Instead, use Anaconda software by opening Anaconda Navigator or Anaconda Prompt from the Start Menu.

Files left behind after uninstalling Anaconda on Windows#

Cause

Some users may need to keep settings files and other users may need to delete them, so Anaconda leaves some settings files in place when it is uninstalled. Specifically, the directories .spyder2, .ipython, .matplotlib, and .astropy remain. Depending on your version of Windows, these may be in C:\Documents and Settings\<USERNAME> or in C:\Users\<USERNAME>.

Note

Replace <USERNAME> with your Windows username as it appears in the Documents and Settings or Users folder.

Solution

Manually delete any unneeded settings files.

Spyder errors or failure to launch on Windows#

Cause

This may be caused by errors in the Spyder setting and configuration files.

Solution

Close and relaunch Spyder and see if the problem remains.
Open the Start Menu and then click Reset Spyder Settings and see if the problem remains.
Close Spyder and relaunch it from Anaconda Prompt with the following command:
```
spyder
```
Delete the directory .spyder2 and then repeat the first item in this list. Depending on your version of Windows, .spyder2 may be in C:\Documents and Settings\<USERNAME> or in C:\Users\<USERNAME>.

Note

Replace <USERNAME> with your Windows username as it appears in the Documents and Settings or Users folder.

Uninstaller requests admin privileges on Windows#

Cause

After installing Anaconda or Miniconda as a non-administrative user on Windows, uninstalling may prompt for administrative privileges.

This occurs when running the uninstaller by choosing Control Panel > Uninstall a program, selecting Anaconda or Miniconda from the list of programs, and clicking Uninstall.

Solution

Open the anaconda3 or miniconda3 installation folders and run the .exe file uninstaller from that location. Uninstallation will complete without prompting for administrative privileges.

EXAMPLE: If you installed Miniconda3, the uninstall file will be Uninstall-Miniconda3.exe. Users who installed Miniconda2 or Anaconda will find a similar file with the appropriate name.

Windows permission errors when installing from Favorites folder#

Cause

The Windows Favorites folder has unusual permissions and may cause permission errors with installers of any software. If you try launching the installer from the Favorites folder, you may see errors such as “Setup was unable to create the directory”, “Access is denied”, or “Error opening file for writing”.

Solution

Move the installer to a different folder and run the installer from the new folder.

Trouble with activation on PowerShell on Windows#

Some users might run into the following backtrace on Windows:

File "C:\Users\damia\Miniconda3\lib\site-packages\conda\activate.py", line 550, in _replace_prefix_in_path
assert last_idx is not None
AssertionError

Solution

Open a Command Prompt window.
Navigate to where you installed conda. The following command shows the default:
```
cd C:\Users\<USERNAME>
```
Run the following command:
```
python -m conda init
```
Close the Command Prompt window.

If this doesn’t work, try running:

conda update conda

MacOS-specific issues#

.zshrc not updated under macOS Catalina#

Cause

MacOS Catalina changed the default shell from Bash to zsh.

Solution

To initialize conda for zsh:

Open a terminal application and run the following command:
```
bash -c "conda init zsh"
```
Close and reopen your terminal application.

Segmentation fault on package import with macOS Python 3.7 interpreter#

In CPython < 3.8, using python3-config to determine a linking command line to compile an extension module will cause that extension module to segfault upon import. python3-config does provide command-line flags, but for the different purpose of embedding a Python interpreter.

Cause

This is because of the command-line flags returned by python3-config. Before Python 3.8, those are needed to embed the core Python interpreter into a different project altogether and not those that should be used when linking a Python extension module.

Python modules should never link to the core Python interpreter library directly, either statically at build time or dynamically at runtime. This is because the Python executable itself provides all the necessary functions and symbols.

Solution

You should only use python*-config —ldflags when linking to an interpreter library (either static or shared).

Action	Python < 3.8	Python >= 3.8
Get command line to link to extension module	`python -c "import sysconfig; print(sysconfig.get_config_var('LDSHARED'))"`	`python3-config --ldflags`
Get command line to embed Python interpreter	`python3-config --ldflags`	`python3-config --ldflags --embed`

python3-config doesn’t include the command/compiler name, whereas the sysconfig way does. This works provided none of your arguments have spaces:

python -c "import sysconfig; print(' '.join(sysconfig.get_config_var('LDSHARED').split(' ')[1:]))"

Linux-specific issues#

Missing libgomp on Power8#

If the Python command import numpy fails, the system is likely missing the libgomp system library.

Cause

Most Power8 Linux distributions include libgomp, but some may not.

Solution

Check whether the system is missing libgomp by running the following command:

conda inspect linkages -n root numpy

If libgomp.so.1 is listed in the “not found:” section, it must be installed.

Install libgomp on Ubuntu by running the following command:

apt install libgomp1

Install libgomp on Red Hat Enterprise Linux (RHEL) or CentOS by running the following command:

yum install libgomp

Anaconda on Power8 reports “can not execute binary file”#

Cause

Anaconda on Power8 only supports little endian mode. The little endian Python binary will not execute on a big endian operating system.

Solution

Install Anaconda on Power8 on a little endian Linux installation or VM.

Troubleshooting#

General issues#

403 error#

HTTP 000 CONNECTION FAILED#

Using Anaconda behind a firewall or proxy#

InsecurePlatformWarning error#

Anaconda search error: not recognized as an internal or external command/unrecognized arguments#

Conda: command not found on macOS or Linux#

Conda: Channel is unavailable/missing or package itself is missing#

Collecting package metadata (repodata.json): - Killed#

Linking problems when Python extensions are compiled with gcc#

Error message: Unable to remove files#

Using 32- and 64-bit libraries and CONDA_FORCE_32BIT#

Installation issues#

Cannot get conda to run after installing#

Recovering your Anaconda installation#

“This package is incompatible with this version of macOS” error when running a .pkg installer on OSX#

Anaconda installer download problems#

Error message on Miniconda install: Already installed#

Conda update anaconda command does not install the latest version of Anaconda#

“The installation failed” message when running a .pkg installer on OSX#

Windows-specific issues#

Cannot see Anaconda menu shortcuts after installation on Windows#

Windows error: Failed to create Anaconda menus or Failed to add Anaconda to the system PATH#

Anaconda interfering with other software on Windows#

Files left behind after uninstalling Anaconda on Windows#

Spyder errors or failure to launch on Windows#

Uninstaller requests admin privileges on Windows#

Windows permission errors when installing from Favorites folder#

Trouble with activation on PowerShell on Windows#

MacOS-specific issues#

.zshrc not updated under macOS Catalina#

Segmentation fault on package import with macOS Python 3.7 interpreter#

Linux-specific issues#

Missing libgomp on Power8#

Anaconda on Power8 reports “can not execute binary file”#