Bug#1072380: cloud.debian.org: Azure deployment misconfigures /etc/hosts resulting in slow sudo

[Noah Meyerhans]

On Wed, Jun 05, 2024 at 10:09:25PM +0100, Luca Boccassi wrote:
> > Is there any specific additional testing that the systemd maintainers
> > would like to see?
> >
> > noah
> 
> The checks themselves look good to me, but would be good doing the
> same validation on a real machine running stable, not just a VM. Bonus
> points for a container too - running a full image like nspawn or lxc.

Updated testing checklist:

[*] Fresh install of libnss-myhostname (nsswitch.conf lists the modules                                               
    in the expected order)                                                                                            
[*] Upgrade of libnss-myhostname (this does not attempt to rewrite                                                    
    nsswitch.conf, which is the same as upgrading to the fixed version                                                
    in trixie)                                                                                                        
[*] Validate that the name resolution behavior is correct with the new                                                
    nss module ordering; that is attempts to resolve the local hostname,                                              
    localhost.localdomain, _gateway, and _outbound are handled by                                                     
    nss-myhostname and don't result in a DNS lookup                                                                   
[*] Validate that libnss-mymachines resolves local container names
    without a DNS query
[*] Validate that resolution of external names is unimpacted                                                          
[ ] validate that a cloud image build based on the updated packages                                                   
    lists the nss modules in the desired order, with myhostname ahead                                                 
    of dns                                                                                                            

To improve readability, I've documented the actual test results in HTML
at https://people.debian.org/~noahm/bug-1072380-testing.html  It
includes tests on bare-metal bookworm hosts as well as systemd-nspawn
bookworm containers.

I haven't yet started validated that the right thing happens when
building a VM image from scratch, which is necessary to confirm the fix
in the environment where the issue was reported, but I think it's
reasonable to initiate the conversation with the SRMs with that test
still pending.

Agree?

noah