Bugs¶

Report a Bug¶

Before you report a bug, please make sure you are using the latest nscli version. Check for available upgrades with nscli -v. If you are having trouble upgrading from one version of NSCLI to another, please try uninstalling NSCLI and performing a clean install. See Installation for instructions.

Please visit our github to submit an issue about NSCLI, or see the Contact section for other ways to get in touch.

Also try reviewing our NSCLI troubleshooting wiki page for additional tips. NSCLI - Issues / Bugs / Known Workarounds

Known Bugs¶

RSA Tokens¶

Moving to, or set up, of the RSA App on your phone requires an additional step to support NSCLI. You are required to set up an additional PIN that is used in place of a token during RSA prompts in NSCLI. Please follow these instructions for set up. Instructions

nscli maint port-migrate¶

(Bug) If switch has a directly connected Aggr and a chained switch to other Aggr, not directly connected, second aggr may be missed
(BUG) Fix finding mlag/vpc if necessary ‘mlag <po #>’ and ‘vpc <po #>’ - only applies to devices with port-channels

Reset¶

NSCLI maintains a cache of credentials and other information. nscli reset will delete these things, and is often enough to fix what looks like buggy behavior. nscli reset is often a good first step in troubleshooting any nscli problems.

Settings¶

There is an nscli settings file that contains usernames and preferences. Sometimes bug-like behavior is caused by missing or incorrect data in the settings file. To see the location of the file, run nscli --version. If the file is deleted, nscli will replace it with a fresh, blank copy on the next run.

Unicode¶

Internally, Python stores text as unicode code points. When it’s time to print some output, Python takes cues from the output device on what types of encodings it will accept. If nscli is running in a terminal, and that terminal wants ascii output, then nscli will try to oblige. The problem is that many unicode characters can’t be properly represented in ascii, so if Python tries to decode one of these characters as ascii, you’ll see a crash with a message about a “UnicodeDecodeError”. Unicode characters are ubiquitous at Rackspace (and in the world). We have them in account names, customer contact names, device management guidelines, etc., so failing to support/encode/display unicode is not an option.

If nscli warns that stdout or stderr lack unicode support, you will need to configure your environment to permit Python to print unicode to the terminal. The easiest way to do this is to set your locale environment variable properly. Run locale -a to see which locales are supported on your system. Then edit your shell profile so that this variable will be set whenever you start a new session. For example:

printf '\nLC_CTYPE=en_US.UTF-8; export LC_CTYPE\n' >> ~/.bash_profile

If nscli does not give a warning regarding unicode, but its unicode output is using the wrong symbols (commonly a diamond or square around a question mark), then nscli is successfully writing unicode to the terminal, but the terminal application itself cannot display the given code point. This problem is outside of nscli’s control, and might be solved by upgrading, replacing, or configuring the terminal application. Some users have had success with choosing a different font.

Illegal instruction: 4¶

Mac users who encounter this error message should upgrade their OS to at least version 10.13. Details available at https://github.rackspace.com/NSI/nscli/issues/1037

Unable to connect to agent¶

If nscli ssh or another nscli command that involves SSHing crashes, and the error output includes Unable to connect to agent, first see the ssh-key documentation in the nscli Installation article.

If you have configured your ssh keys/agent properly, but you are still seeing this crash, and you are running Ubuntu 18, see this workaround: https://github.com/Microsoft/WSL/issues/3183#issuecomment-411138426

Reaching Into NSCLI¶

In rare cases, it may be useful to reach into the internals of NSCLI’s virtual environment and take some action, for example, upgrading one of NSCLI’s Python dependencies to apply a critical patch. NSCLI’s python interpreter is shown in the output of nscli -v, therefore its pip executable can be invoked like this: "$(nscli -v 2>&1 | awk '/python interpreter/ {print $4}')" -m pip

Again, it should not normally be necessary to interact with this at all, but for example, if one needed to upgrade NSCLI’s NSPYLIB dependency, it could be done like this:

"$(nscli -v 2>&1 | awk '/python interpreter/ {print $4}')" -m pip install --upgrade nspylib

Binding to AD¶

The error message Unable to determine user's division. indicates that NSCLI was not able to get the information it needs from Active Directory via the Identity API. This could mean that your SSO username or identity token are wrong in NSCLI, which are problems that an nscli reset will fix. It could also mean that api.identity.rackspace.corp is not reachable from your system, which could indicate a service outage or a local networking problem.

PTY Allocation Failed¶

The PTY allocation request failed on channel 0 error message happens because the user is not a member of the Balabit_Support_gateway LDAP group (or the corresponding group for non-support balabits). Despite this, the balabit still allows the users to complete authentication; it only fails later in the process when trying to allocate a pseudo-terminal.

ModuleNotFoundError: No module named ‘distutils.util’¶

The ModuleNotFoundError: No module named 'distutils.util' error message happens on Linux because the python3 distutils package is not installed. To install the module run: sudo apt-get install python3-distutils

nscli maint port-audit¶

port-audit attempts to gather a device’s switch information from CTK to see which DC it is in in case of a DC migration. This information is no longer being populated for new switches and the script will now fall back to using the device’s DC instead. If a device is going through a migration to a new DC, the results may be unreliable or just wrong.