samba - Nov 2015 - Authentication to Secondary Domain Controller initially fails when PDC is offline

Hi Ole,

I'm still not answering your issue but I come back to speak about TTL.
Perhaps someone would be able to bring us some light on that.

This morning I'm trying to reproduce the way I do broke my test AD domain.
This leads me to deal with SOA record (I broke my test AD seizing FSMO
roles before removing old FSMO owner, SOA was not changed during that
process and I suspect this was one of the point leading to all issues this
test domain has)

Anyway:
samba-tool dns query m700 samba.domain.tld samba.domain.tld SOA -k yes
  Name=, Records=1, Children=0
    SOA: serial=1, refresh=900, retry=600, expire=86400, *minttl=3600*,
ns=m700.samba.domain.tld., email=hostmaster.samba.domain.tld.
(flags=600000f0, serial=1, *ttl=3600*)
  Name=_msdcs, Records=0, Children=0
  Name=_sites, Records=0, Children=1
  Name=_tcp, Records=0, Children=4
  Name=_udp, Records=0, Children=2
  Name=DomainDnsZones, Records=0, Children=2
  Name=ForestDnsZones, Records=0, Children=2
  Name=m700, Records=0, Children=0

This shows us TTL is in fact equal to minimumttl inside AD DB.

According to
http://stackoverflow.com/questions/20297531/meaning-of-the-five-fields-of-the-answer-section-in-dig-query
the second member of dig's answer section is TTL.

dig -t soa samba.domain.tld
...
samba.domain.tld. *3593* IN      SOA     m700.samba.domain.tld.
hostmaster.samba.domain.tld. 1 900 600 86400 3600
...
 When yesterday the same request gave the following answer:

...
samba.domain.tld. *1715* IN      SOA     DC1.samba.domain.tld. 62 900 600
86400 3600
...

So I ran several that same command and each the value displayed as second
member (here 1715 or 3593) was descreased by the same amount of second as
the time between my command launchs.

It seems this shown TTL is declared TTL (or minttl) minus the amount of
seconds since last renewal of this TTL. No idae why this behaviour. If
someone knows, I would be pleased to learn :)

Cheers,

mathias

2015-11-20 8:44 GMT+01:00 Mueller <mueller at tropenklinik.de>:
> Within a real windows 2008 Domain it ist he same behaviour. Even there you
> need the clients to reboot.
> This feature got lost after the beta status of samba 4. I had a test
> environment with the first betas and it worked there without any issue.
> Even mapping the shares by domain
> worked: \\my.domain\share.  Test this with the new versions, it will fail.
> (only netlogon will do).
> Would be fine samba does it better. Samba 4 ist to close building windows
> AD server. It would be a great step it adds his own better features.
>
> Greetings
>
> Daniel
>
>
> EDV Daniel Müller
>
> Leitung EDV
> Tropenklinik Paul-Lechler-Krankenhaus
> Paul-Lechler-Str. 24
> 72076 Tübingen
> Tel.: 07071/206-463, Fax: 07071/206-499
> eMail: mueller at tropenklinik.de
> Internet: www.tropenklinik.de
>
>
>
>
> -----Ursprüngliche Nachricht-----
> Von: Ole Traupe [mailto:ole.traupe at tu-berlin.de]
> Gesendet: Donnerstag, 19. November 2015 16:26
> An: mathias dufresne <infractory at gmail.com>
> Cc: samba <samba at lists.samba.org>
> Betreff: Re: [Samba] Authentication to Secondary Domain Controller
> initially fails when PDC is offline
>
> Ok, I see. Nevertheless, thank you very much for your effort!
>
> I must say that I can't actually believe that no one knows an answer to
> this problem. It must affect MANY people using Samba DCs. According to all
> the tests on the wiki, everything is working fine. Then I pull the plug on
> my first DC and no one can log on. And this time I waited far longer than
> the suggested "refresh interval" of 15 min - even longer than the
value
> called "TTL" in the GUI of 1h. I also tried "ipconfig
/flushdns" on my
> windows client. Does not improve the situation. Only a reboot solves the
> issue. But that would be no acceptable practice for Linux member servers.
> And it doesn't seem to help, anyway (just tried this).
>
> It is one of the first and most important tests for a domain to see what
> happens if the first DC is down. Without a working take-over, other DCs are
> nothing more than backup (replication) targets, and the domain is not
> fail-safe.
>
> This can't be the end of the story, right?
>
> Ole
>
>
>
> Am 19.11.2015 um 14:04 schrieb mathias dufresne:
> > No idea about your main issue, I was merely answering to your last
> > question about changing SOA record.
> >
> > Here is another view of that command:
> > samba-tool dns update <server> <zone> <name> SOA \
'OLDnameserver
> > email serial refresh retry expire minimumttl' \ 'NEWnameserver
email
> > serial refresh retry expire minimumttl'
> >
> > I'm not too confident with DNS internals so I'm not sure if
the TTL
> > you mentioned is or isn't "expire" or
"minimumttl".
> >
> > After digging a little bit it seems previous line is completely wrong,
> > neither "expire" nor "minimumttl" are
"TTL".
> > This because :
> > dig -t SOA SAMBADOMAIN.TLD
> > ...
> > samba.domain.tld. 1715 IN      SOA DC1.samba.domain.tld. 62 900 600
> > 86400 3600
> > ...
> >
> > And from what I just read in dig "ANSWER SECTION" the second
field is
> > the TTL, so 1715 in my case, which as nothing to do with
"expire"
> > (86400) or "minimumtll" (3600).
> >
> > And that makes me wondering how TTL can be less than
"minimumttl"...
> >
> > So, the short way: the command I gave do not seem to be designed to
> > help you changing TTL. Sorry : )
> >
> > Cheers,
> >
> > mathias
> >
> > 2015-11-19 13:43 GMT+01:00 Ole Traupe <ole.traupe at tu-berlin.de
> > <mailto:ole.traupe at tu-berlin.de>>:
> >
> >     Mathias, thank you very much for your comprehensive instructions!
> >
> >     Just one question: Harry suggested that, in order to overcome the
> >     below DNS related problems, the TTL would have to be adjusted
> >     (lowered). However, the TTL seems to be the only time value not
> >     covered by the command provided by you.
> >
> >     Is it really the TTL that is the culprit or is it rather the first
> >     time value (something like "Refresh value" in english)?
> >
> >     Do you know this?
> >
> >     Ole
> >
> >
> >
> >     Am 19.11.2015 um 11:19 schrieb mathias dufresne:
> >>     Hi Ole,
> >>
> >>     You want to change SOA of your AD domain?
> >>
> >>     Here some working command:
> >>     samba-tool dns update <working DC> samba.domain.tld \
> >>     samba.domain.tld SOA \
> >>     'oldSOA.samba.domain.tld. hostmaster.samba.domain.tld. 58
900 600
> >>     86400 3600' \
> >>     'newSOA.samba.domain.tld. hostmaster.saba.domain.tld. 59
900 600
> >>     86400 3600' -k yes
> >>
> >>     This needs you performed some kinit before using an account
able
> >>     to modify this entry (by default only administrator is able to
> >>     that I expect).
> >>
> >>     This must be done for the two DNS zones of your domain:
> >>     samba.domain.tld + _msdcs.samba.domain.tld
> >>
> >>     First number of replacement record (here "59") is
serial number.
> >>     Replication of change seemed to work without changing that
serial
> >>     number but as DNS love to rely on it, changing that serial
should
> >>     be a good idea.
> >>
> >>     Hoping this helps...
> >>
> >>     Cheers,
> >>
> >>     mathias
> >>
> >>
> >>     2015-11-18 16:44 GMT+01:00 Ole Traupe <ole.traupe at
tu-berlin.de
> >>     <mailto:ole.traupe at tu-berlin.de>>:
> >>
> >>
> >>             It is DNS related.
> >>
> >>                 What is the best way of dealing with this?
> >>
> >>             The *best way* is a HA solution for your DNS Servers,
but
> >>             its expensive.
> >>
> >>             The DNS client (resolver) caches the srv records for
15
> >>             minutes aka 900
> >>             seconds.
> >>
> >>             ipconfig /flushdns drops the cache. Reboot does the
same.
> >>
> >>             On server side you may set shorter TTL for the server
> >>             records, but then
> >>             you have more DNS traffic. On small netwoks (sites up
to
> >>             20 clients, no
> >>             wifi) I have good experience with a TTL of 180.
> >>
> >>
> >>         Harry, I tried this - unsuccessfully.
> >>
> >>         I have TTL settings in a) the SOA and b) the NS record of
the
> >>         FQDN and the _msdcs.FQDN sections in my Windows RSAT DNS
> >>         console. None of these 4 entries I can change: I get
> >>         something like "The Source Of Authority (SOA) cannot
be
> >>         updated. The record already exists."
> >>
> >>         Do you have an idea how to accomplish this? Currently the
> >>         setting is 1h, which is pretty long.
> >>
> >>         Ole
> >>
> >>
> >>
> >>
> >>         --
> >>         To unsubscribe from this list go to the following URL and
> >>         read the
> >>         instructions:
https://lists.samba.org/mailman/options/samba
> >>
> >>
> >
> >
>
> --
> To unsubscribe from this list go to the following URL and read the
> instructions:  https://lists.samba.org/mailman/options/samba
>
>

Am 20.11.2015 um 11:54 schrieb mathias dufresne:
> Hi Ole,
>
> I'm still not answering your issue but I come back to speak about TTL. 
> Perhaps someone would be able to bring us some light on that.
>
> This morning I'm trying to reproduce the way I do broke my test AD 
> domain. This leads me to deal with SOA record (I broke my test AD 
> seizing FSMO roles before removing old FSMO owner, SOA was not changed 
> during that process and I suspect this was one of the point leading to 
> all issues this test domain has)
>
> Anyway:
> samba-tool dns query m700 samba.domain.tld samba.domain.tld SOA -k yes
>   Name=, Records=1, Children=0
>     SOA: serial=1, refresh=900, retry=600, expire=86400, 
> *minttl=3600*, ns=m700.samba.domain.tld., 
> email=hostmaster.samba.domain.tld. (flags=600000f0, serial=1, *ttl=3600*)
>   Name=_msdcs, Records=0, Children=0
>   Name=_sites, Records=0, Children=1
>   Name=_tcp, Records=0, Children=4
>   Name=_udp, Records=0, Children=2
>   Name=DomainDnsZones, Records=0, Children=2
>   Name=ForestDnsZones, Records=0, Children=2
>   Name=m700, Records=0, Children=0
>
> This shows us TTL is in fact equal to minimumttl inside AD DB.
Not for me:

SOA: serial=29, refresh=180, retry=600, expire=86400, minttl=180, 
ns=DC2.my.domain.tld., email=hostmaster.my.domain.tld. (flags=600000f0, 
serial=0, ttl=3600)

>
> According to 
>
http://stackoverflow.com/questions/20297531/meaning-of-the-five-fields-of-the-answer-section-in-dig-query
> the second member of dig's answer section is TTL.
>
> dig -t soa samba.domain.tld
> ...
> samba.domain.tld. *3593* IN      SOA m700.samba.domain.tld. 
> hostmaster.samba.domain.tld. 1 900 600 86400 3600
> ...
>  When yesterday the same request gave the following answer:
>
> ...
> samba.domain.tld. *1715* IN      SOA DC1.samba.domain.tld. 62 900 600 
> 86400 3600
> ...
>
> So I ran several that same command and each the value displayed as 
> second member (here 1715 or 3593) was descreased by the same amount of 
> second as the time between my command launchs.
>
> It seems this shown TTL is declared TTL (or minttl) minus the amount 
> of seconds since last renewal of this TTL. No idae why this behaviour. 
> If someone knows, I would be pleased to learn :)
Yes, I thought so. This is "remaining TTL" for you.

Interestingly, for me this value is always constant and equals 1h, no 
matter what.


ANYWAYS, I would like to approach from a different direction:

If my first DC is offline, a ping on any of my domain machines takes 5+ 
seconds to resolve. I figure that my logon problems reflect multiple 
such timeouts during the logon process accumulating to a total duration 
not accepted by the unix logon mechanism.

If there would be ANY way to reduce the time (to 1 s or something) a 
machines waits until it finally accepts that a DNS server just won't 
respond and goes over to the next one... - that actually might solve the 
issue.

Is there an option for this on unix machines?

Ole

On 11/20/2015 7:40 AM, Ole Traupe wrote:
>
>
> Am 20.11.2015 um 11:54 schrieb mathias dufresne:
>> Hi Ole,
>>
>> I'm still not answering your issue but I come back to speak about 
>> TTL. Perhaps someone would be able to bring us some light on that.
>>
>> This morning I'm trying to reproduce the way I do broke my test AD 
>> domain. This leads me to deal with SOA record (I broke my test AD 
>> seizing FSMO roles before removing old FSMO owner, SOA was not 
>> changed during that process and I suspect this was one of the point 
>> leading to all issues this test domain has)
>>
>> Anyway:
>> samba-tool dns query m700 samba.domain.tld samba.domain.tld SOA -k yes
>>   Name=, Records=1, Children=0
>>     SOA: serial=1, refresh=900, retry=600, expire=86400, 
>> *minttl=3600*, ns=m700.samba.domain.tld., 
>> email=hostmaster.samba.domain.tld. (flags=600000f0, serial=1, 
>> *ttl=3600*)
>>   Name=_msdcs, Records=0, Children=0
>>   Name=_sites, Records=0, Children=1
>>   Name=_tcp, Records=0, Children=4
>>   Name=_udp, Records=0, Children=2
>>   Name=DomainDnsZones, Records=0, Children=2
>>   Name=ForestDnsZones, Records=0, Children=2
>>   Name=m700, Records=0, Children=0
>>
>> This shows us TTL is in fact equal to minimumttl inside AD DB.
>
> Not for me:
>
> SOA: serial=29, refresh=180, retry=600, expire=86400, minttl=180, 
> ns=DC2.my.domain.tld., email=hostmaster.my.domain.tld. 
> (flags=600000f0, serial=0, ttl=3600)
>
>
>>
>> According to 
>>
http://stackoverflow.com/questions/20297531/meaning-of-the-five-fields-of-the-answer-section-in-dig-query
>> the second member of dig's answer section is TTL.
>>
>> dig -t soa samba.domain.tld
>> ...
>> samba.domain.tld. *3593* IN      SOA m700.samba.domain.tld. 
>> hostmaster.samba.domain.tld. 1 900 600 86400 3600
>> ...
>>  When yesterday the same request gave the following answer:
>>
>> ...
>> samba.domain.tld. *1715* IN      SOA DC1.samba.domain.tld. 62 900 600 
>> 86400 3600
>> ...
>>
>> So I ran several that same command and each the value displayed as 
>> second member (here 1715 or 3593) was descreased by the same amount 
>> of second as the time between my command launchs.
>>
>> It seems this shown TTL is declared TTL (or minttl) minus the amount 
>> of seconds since last renewal of this TTL. No idae why this 
>> behaviour. If someone knows, I would be pleased to learn :)
>
> Yes, I thought so. This is "remaining TTL" for you.
>
> Interestingly, for me this value is always constant and equals 1h, no 
> matter what.
>
>
> ANYWAYS, I would like to approach from a different direction:
>
> If my first DC is offline, a ping on any of my domain machines takes 
> 5+ seconds to resolve. I figure that my logon problems reflect 
> multiple such timeouts during the logon process accumulating to a 
> total duration not accepted by the unix logon mechanism.
>
> If there would be ANY way to reduce the time (to 1 s or something) a 
> machines waits until it finally accepts that a DNS server just won't 
> respond and goes over to the next one... - that actually might solve 
> the issue.
>
> Is there an option for this on unix machines?
>
> Ole
You can add your DC's to your hosts file. Usually your hosts file is 
queried first, prior to DNS for resolve.

One thing I notice a bit odd is this

SOA: serial=29, refresh=180, retry=600, expire=86400, minttl=180, 
*ns=DC2.my.domain.tld.*, email=hostmaster.my.domain.tld. 
(flags=600000f0, serial=0, ttl=3600)

Normally your name server would be the same as your DC who is SOA. Did 
you manually change this from DC1 to DC2? What DC is your SOA?







-- 
-James