On 4/24/08 Jerry wrote:
> You are tracing the client. But the log only shows the
> parent winbindd process. I would check the child
> processes because I'll bet you have more traffic that
> will illuminate what is going on in those logs.
Thanks for the tip. I took your advice and ran 'strace -ff' on winbind
and found the problem. It was trying to use mDNS to locate the kdc.
However, our domain is unicast and uses the .local extension. I added
the line 'mdns off' to the /etc/host.conf file (apparently it defaults
to on) and it eliminated the 30 second timeout pause :)
Here's the line in the strace output that tipped me off:
17:24:34 sendto(20,
"\241q\1\0\0\1\0\0\0\0\0\0\20_kerberos-master\4_u"..., 54, 0,
{sa_family=AF_INET, sin_port=htons(5353),
sin_addr=inet_addr("224.0.0.251")}, 28) = 54
17:24:34 poll([{fd=20, events=POLLIN}], 1, 5000) = 0
This poll() call is what was actually timing out. The timeout was 5s and
it did this multiple times. Now that mDNS is turned off it makes this
request directly to the kdc rather than trying to search for it.
WoooHooo!
Thanks,
Matt
--
To unsubscribe from this list go to the following URL and read the
instructions: https://lists.samba.org/mailman/listinfo/samba
