On 11/23/18 11:03 AM, Harald Dunkel wrote:
On 11/22/18 2:55 PM, Michael Ströder wrote:
> On 11/22/18 2:23 PM, Harald Dunkel wrote:
>> I am not sure if waiting for the interface solves the problem. It
>> should wait for DNS to succeed, shouldn't it?
>> I would prefer if the backend waits for DNS a little bit longer, of
> Does the backend stay dead after /etc/resolv.conf being corrected?
Yes. I have to restart sssd to make it work. If I don't, then sssd_nss
Hmm, AFAIK sssd aims at providing NSS/PAM services also on laptops where
the DNS config might change every now and then. Therefore I'd consider
this to be a bug anyway.
>> Surely systemctl status sssd should not say
"running", while the
>> backend is dead.
> Not sure about that. The NSS and PAM responders serve cached data from
> the local DB even in the case all backends are marked dead. Of course
> this does not help you if there's no cached data available yet.
I understand that the local database is an important part of sssd,
but it should be possible to separate accessing the network services
from providing cached data.
Disclaimer: I'm not involved in sssd development.
It's a single process started by systemd. Your wish would require much
work distinguishing various cases of failure during startup. Therefore
I'd argue that fixing the reconnection bug is better than putting effort