winbindd stuck at getaddrinfo.

[Hemanth Thummala]

Hi Richard,

In our case, nmbd was running fine. We have verified the daemon status and
it was running.

Patching getaddrinfo with "poll instead of kevent" was also not helpful.
Would like to know if heimdal version of getaddrinfo() will actually
address this issue. I am trying to compile these changes right now.

In general getaddrinfo() doesn't have any timeout option. Having timeout in
resolv.conf is only way to configure the timeout for DNS responses. But the
results are inconsistent for me with these resolv.conf changes.

Thanks,
Hemanth.

On Thu, Oct 30, 2014 at 10:04 AM, Richard Sharpe <
realrichardsharpe at gmail.com> wrote:

> On Thu, Oct 30, 2014 at 7:56 AM, Richard Sharpe
> <realrichardsharpe at gmail.com> wrote:
> > On Wed, Oct 29, 2014 at 11:14 AM, Richard Sharpe
> > <realrichardsharpe at gmail.com> wrote:
> >> On Wed, Oct 29, 2014 at 12:20 AM, Hemanth Thummala
> >> <hemanth.thummala at gmail.com> wrote:
> >>> Hi All,
> >>>
> >>> We are using samba 3.6.12+ stack on FreeBSD 8.0. Recently we are
> hitting a
> >>> winbindd hung issue at few of our customers. Each time we could see
> that
> >>> winbindd getting hung at gettaddrinfo.
> >>
> >> Curiously, I am seeing what might be a similar issue at one site in
> >> Italy. In this case it manifests as a long pause during
> >> authentication, which looks like a winbindd timeout in
> >> wb_is_trusted_domain, but because none of the functions in
> >> lib/winbind_util.c has any DEBUG statements in them I can only guess
> >> that this is the case. The site has a very large domain (it is part of
> >> a Class A network that is not 10.0.0.0.)
> >>
> >> This is with 3.6.9 (CentOS 6.x).
> >>
> >> I have provided them with modified RPM to give me extra info at log
> >> level 10 and should know more later today.
> >>
> >> I will look at the master code in this area and might submit a patch
> >> that provides extra info at log level 10 because that would sure
> >> improve our ability to debug things quickly.
> >
> > Our problem seems to have been that nmbd was not running. Now to
> > figure out why winbindd might timeout or return another error during
> > auth if nmbd is not running.
>
> Hmmm, it seems that RedHat switched from Samba 3.5.10 to 3.6.9 between
> RHEL 6.3 and 6.4, so the problem I was seeing was likely on Samba
> 3.5.10, not on 3.6.x.
>
> --
> Regards,
> Richard Sharpe
> (何以解憂？唯有杜康。--曹操)
>