Fedora/CentOS: Failed at step NAMESPACE spawning /usr/bin/tor: No such file or directory

[nusenu]

Errors:

CentOS 7, tor 0.2.9.16

systemd: Starting Anonymizing overlay network for TCP (instance: 10.0.2.15_9000)...
systemd: Failed at step NAMESPACE spawning /usr/bin/tor: No such file or directory
systemd: [email protected]_9000.service: control process exited, code=exited status=226
systemd: Failed to start Anonymizing overlay network for TCP (instance: 10.0.2.15_9000).
systemd: Unit [email protected]_9000.service entered failed state.
systemd: [email protected]_9000.service failed.

Fedora 28, tor 0.3.2.10

[email protected]_9000.service: Failed to set up mount namespacing: No such file or directory
[email protected]_9000.service: Failed at step NAMESPACE spawning /usr/bin/tor: No such file or directory

[nusenu]

https://bugzilla.redhat.com/show_bug.cgi?id=1633568

[nusenu]

I've been in contact with the RPM maintainer and he will get to this eventually (at some point in 2019)

[ladar]

I wrote a shell script for setting up a TOR relay on CentOS/RHEL 8. At the moment its only for Guard/Middle relays.

The script also contains my first attempt at tuning a system, in order to get the most out of TOR.

To use it, simply run the centos-8-tor.sh script file. The script will try to auto detect your public IP, and pick a random nickname, but both can be set manually using environment variables.

The attached tarball contains some additional non-TOR specific system setup scripts. These scripts setup generic services like configuring NTP, fail2ban, automatic updates, etc. To use it, simply decompress the tarball and run the centos-8-base.sh instead. It will automatically run the TOR script at the end.

centos-8-tor.sh.txt
centos-8-complete.tar.gz

[nusenu]

Hi ladar,
since you are posting it in this issue about the NAMESPACE error when using the multi-instance systemd service file: Did you solve that problem in your work?

[ladar]

@nusenu the script I uploaded was built around CentOS 8, which has different issues. In retrospect I probably should have posted my comment on issue #209.

For CentOS 8, the Official TOR package actually fails to run (it's linked against a library that isn't available), but the EPEL package works just fine, and is the latest official release. If you look at my script, it sets up the official TOR repo, but disables the repo, so the script ends up installing the EPEL version.

If you're using CentOS 7, the reverse is true. The EPEL TOR package is out of date (0.3.5 as of this writing), but the version in the RPM in the official TOR repo is current, and works perfectly.

I can upload a CentOS 7 version of the script if people are interested. That script would use drastically different tuning/kernel params, since the CentOS 7 kernel doesn't support/use the same kernel parameters. The bbr scheduler springs to mind.

P.S. Note the script I uploaded has a few typos. But the most important is that when it dumps out the /etc/tor/torrc config file, it uses the variable HICKNAME instead of NICKNAME. Since HICKNAME isn't set, the field ends up empty, and the relay is listed as unnamed.

[duritong]

EPEL 7 is on the latest Tor LTS release: https://trac.torproject.org/projects/tor/wiki/org/teams/NetworkTeam/CoreTorReleases this is something different than "being out of date".

[ladar]

@duritong my apologies. I thought a relay with v0.3.5.10 would get flagged for having an out-of-date version, but I was mistaken.

[nusenu]

@nusenu the script I uploaded was built around CentOS 8, which has different issues. In retrospect I probably should have posted my comment on issue #209.

I'm primarily

For CentOS 8, the Official TOR

"It's Tor, not TOR. Pass it on." https://twitter.com/torproject/status/1301583526765051905

package actually fails to run (it's linked against a library that isn't available)

Did you report this bug to the torproject?

[nusenu]

I'm closing this since CentOS is discontinued.

[duritong]

Well CentOS Stream 8 is a valid and straight forward replacement for CentOS Linux. Switching to it takes 2 commands and no Ansible adjustments. And then there are Rocky and Alma Linux, which continue to rebuild RHEL.