← Back to team overview

debcrafters-packages team mailing list archive

[Bug 2112595] Re: `Domains=local` is appending `.local` to the FQDN

 

...better late than never xD Thanks for your patience @raof and @enr0n!

** Description changed:

- ==== Bug
+ =================== SRU ===================
  
  The latest release of gce-compute-packages (a.k.a google-compute-engine)
  20250328.00-0ubuntu2 drops a config file into
  /etc/systemd/resolved.conf.d that adds a `.local` prefix to the FQDN.
  
  ````
  ubuntu@jammy-lp:~$ hostname -f
  jammy-lp.local
  ````
  
  [Impact]
  
  This will likely affect users with scripts calling `hostname` or those
  who rely on a known FQDN
  
- [Mitigation]
+ [Current Mitigation]
  
  Editing the `gce-resolved.conf` file to use `Domains=~local` suffices. A
  one line fix to use is:
  
  ```
  sudo sed --in-place 's,Domains=local,Domains=~local,g' /etc/systemd/resolved.conf.d/gce-resolved.conf
  ```
  
  followed by restarting the `systemd-resolved.service`:
  
  ```
  sudo systemctl restart systemd-resolved.service
  ```
  
  fixes the issue:
  
  ```
  ubuntu@jammy-lp:~$ hostname -f
  jammy-lp.c.ubuntu-kajiya.internal
  ```
+ 
+ [Test Case]
+ 
+ For this regression in particular, to test the fix is quite
+ straightforward. After the new version has been installed, after running
+ `sudo systemctl restart systemd-resolved.service`, `hostname -f` should
+ return the proper FQDN as opposed to the shortname appended with local
+ (i.e. `jammy-lp.c.ubuntu-kajiya.internal` vs. `jammy-lp.local`). For
+ this bug specifically, that's all the testing needed. However, as we are
+ very careful with any new releases of the guest agent pkgs, the
+ following will also happen:
+ 
+  * an image based on -proposed will be built for GCE and published to the ubuntu-os-cloud-image-proposed project
+  * the CPC team will run internal validations (CTF) and Google's upstream test suite `cloud-image-tests` (CIT)
+  * the GCE team will be asked to validate that the new package addresses the issues it is expected to address, and that the image passes their own internal image validation.
+ 
+ If all the testing indicates that the new package is acceptable,
+ verification can be considered done.
+ 
+ Special care will also be taken to ensure that the SRU-ed version also
+ plays nice with the sister SRU at LP: #2113788 as well.
+ 
+ [Vendored Dependencies]
+ 
+ N/A for gce-compute-image-packages
+ 
+ [Where Problems Could Occur]
+ 
+ For this bug specifically the risk is very low. The workaround listed
+ above is already deployed on GCP images and has been for weeks with no
+ reported regressions or issues. The fix implemented here is identical to
+ the work around (just instead of `sed`-ing the file it is completely
+ replaced with the fix instead). Between this information and CPC's own
+ internal test harness (CTF), there is confidence that most if not all
+ "edge cases" and/or obvious regressions concerns can be dismissed before
+ the new version lands in `-updates`.
+ 
+ [Other Information]
+ 
+ This bug is used for tracking of releasing the new upstream version for
+ all supported series, as per the approved policy mentioned in the
+ following MRE:
+ 
+ https://wiki.ubuntu.com/gce-compute-image-packages-Updates
+ 
+ This package is only used on AMD64 and ARM64 but is built for all
+ available architectures.

** Summary changed:

- `Domains=local` is appending `.local` to the FQDN
+ [SRU] `Domains=local` is appending `.local` to the FQDN

-- 
You received this bug notification because you are a member of
Debcrafters packages, which is subscribed to gce-compute-image-packages
in Ubuntu.
https://bugs.launchpad.net/bugs/2112595

Title:
  [SRU] `Domains=local` is appending `.local` to the FQDN

Status in gce-compute-image-packages package in Ubuntu:
  Fix Released

Bug description:
  =================== SRU ===================

  The latest release of gce-compute-packages (a.k.a google-compute-
  engine) 20250328.00-0ubuntu2 drops a config file into
  /etc/systemd/resolved.conf.d that adds a `.local` prefix to the FQDN.

  ````
  ubuntu@jammy-lp:~$ hostname -f
  jammy-lp.local
  ````

  [Impact]

  This will likely affect users with scripts calling `hostname` or those
  who rely on a known FQDN

  [Current Mitigation]

  Editing the `gce-resolved.conf` file to use `Domains=~local` suffices.
  A one line fix to use is:

  ```
  sudo sed --in-place 's,Domains=local,Domains=~local,g' /etc/systemd/resolved.conf.d/gce-resolved.conf
  ```

  followed by restarting the `systemd-resolved.service`:

  ```
  sudo systemctl restart systemd-resolved.service
  ```

  fixes the issue:

  ```
  ubuntu@jammy-lp:~$ hostname -f
  jammy-lp.c.ubuntu-kajiya.internal
  ```

  [Test Case]

  For this regression in particular, to test the fix is quite
  straightforward. After the new version has been installed, after
  running `sudo systemctl restart systemd-resolved.service`, `hostname
  -f` should return the proper FQDN as opposed to the shortname appended
  with local (i.e. `jammy-lp.c.ubuntu-kajiya.internal` vs. `jammy-
  lp.local`). For this bug specifically, that's all the testing needed.
  However, as we are very careful with any new releases of the guest
  agent pkgs, the following will also happen:

   * an image based on -proposed will be built for GCE and published to the ubuntu-os-cloud-image-proposed project
   * the CPC team will run internal validations (CTF) and Google's upstream test suite `cloud-image-tests` (CIT)
   * the GCE team will be asked to validate that the new package addresses the issues it is expected to address, and that the image passes their own internal image validation.

  If all the testing indicates that the new package is acceptable,
  verification can be considered done.

  Special care will also be taken to ensure that the SRU-ed version also
  plays nice with the sister SRU at LP: #2113788 as well.

  [Vendored Dependencies]

  N/A for gce-compute-image-packages

  [Where Problems Could Occur]

  For this bug specifically the risk is very low. The workaround listed
  above is already deployed on GCP images and has been for weeks with no
  reported regressions or issues. The fix implemented here is identical
  to the work around (just instead of `sed`-ing the file it is
  completely replaced with the fix instead). Between this information
  and CPC's own internal test harness (CTF), there is confidence that
  most if not all "edge cases" and/or obvious regressions concerns can
  be dismissed before the new version lands in `-updates`.

  [Other Information]

  This bug is used for tracking of releasing the new upstream version
  for all supported series, as per the approved policy mentioned in the
  following MRE:

  https://wiki.ubuntu.com/gce-compute-image-packages-Updates

  This package is only used on AMD64 and ARM64 but is built for all
  available architectures.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/gce-compute-image-packages/+bug/2112595/+subscriptions



References