After making a clean install of Buster and setup it, the system doesn't
boot propery and enters emergency mode with some systemd-udevd errors on
timing out.

I tracked down and isolated the issue to be caused by nss-ldap group
mapping: If I remove ldap from nsswtich.conf groups (only for groups table)
the system boots fine (So I can use ldap for everything else except for
group)

I already faced the same problem long time ago (not sure, but I think on
jessie and ubuntu older releases) and the workarround/solution was to set
nss_init_groups_ignore users to list all localacounts (so don't lookup for
LDAP groups for local accounts). This time this didn't worked, as I updated
my nss_init_groups_ignore_users to the list of current local users with no
luck.

Some details tested/discarded:
- there are no custom udev rules making use of any LDAP user/group
- I tried setting various timeouts/soft_policy on LDAP configuration
- Also tried [UNAVAIL=return] and other similar optons on nsswitch.conf
- Exactly the same configuration works perfectly on Debain Stretch (as I
configure them thru Ansible)

Note that while researching I found many similar bug reports (also on
different distros) related to this issue, all of them providing
workarrounds (which didn't worked) but none providing a permanent FIX or
solution:

https://bugs.launchpad.net/ubuntu/+source/libnss-ldap/+bug/1024475
https://bugs.launchpad.net/ubuntu/+source/libnss-ldap/+bug/51315
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=318622
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=339797
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=349509
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=375077
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=375215
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=388729
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=391167
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=441458
https://bugzilla.redhat.com/show_bug.cgi?id=234541
https://bugzilla.redhat.com/show_bug.cgi?id=187852

So basically seems that NSS-LDAP is queried when is still not ready, either
because LDAP server is down or the affected system didn't initialized
network yet. Either the case, this shouldn't prevent the system to boot
normally. More than a bug on libnss-ldap/udev seems a wrong/unstable
integration on the init process. I don't know if any NSS network servicces
(NIS?, winbind?) experience similar issues or how they avoid them.

Does any one faced same issue or can provide any help/workarround/clues?
Should I open a new bug report?

Thanks much for any hint/help

Reply via email to