Networker

Re: [Networker] Client Addition bug in 6.1.3

2003-07-15 15:57:43
Subject: Re: [Networker] Client Addition bug in 6.1.3
From: "Mark Bradshaw (BTOpenWorld)" <notthehoople AT BTOPENWORLD DOT COM>
To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
Date: Tue, 15 Jul 2003 21:12:14 +0100
I've seen this problem on a Solaris 8 server which had a slightly broken DNS
setup which didn't look, at first glance, to be broken at all!

Basically what the problem was was that there were multiple DNS servers
listed under resolv.conf but these nameservers had a different idea of what
the world looked like. The backup clients I was adding existed in one of the
DNS servers but not the other. This cause NetWorker a problem - NetWorker
would sometimes fail to add the client successfully, other times everything
worked ok. Nslookup appeared to work correctly as it tried the nameservers
in turn until it got a good answer. NetWorker didn't appear to wait long
enough to get the answer if the first nameserver didn't know how to resolve
the client name.

Another symptom of this was that a client would sometimes add successfully
but fail to backup with the error message "not a registered client".

Once the name and IP address was in the backup server's cache (i.e. Through
a ping) then NetWorker worked fine and added the client. This was only
tracked down by carefully watching what nslookup was telling us about each
client and noting which DNS server was replying.

The Backup Server had been in position for sometime and had worked
successfully under 5.x for ages. Upgrading to 6.1.3 brought these issues to
light...

This may not apply to your situation, but thought it might be worth
mentioning!

Cheers

Mark

> It wasn't a problem with my environment until upgrading to 6.x.
>
> Not sure why nslookup is considered unreliable when used to query
> internal DNS entries.  But I'll verify that getent returns the proper
> entries when I have a problem adding my next client.  However, if it was
> a DNS issue then it would stop working again once I removed the
> /etc/hosts entry from my server, but after that initial backup
> everything works perfectly.
>
> Joel
>
>
>
> -----Original Message-----
> From: Kent Holloway [mailto:Kent.Holloway AT ONVOY DOT COM]
> Sent: Friday, July 11, 2003 2:50 PM
> To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
> Subject: Re: [Networker] Client Addition bug in 6.1.3
>
> We have this problem also on 6.1.2 on Solaris 8.
> I have double and triple checked our DNS settings in resolv.conf and
> nsswitch.conf.
> Hosts added have both proper forward and reverse DNS and we need to
> stop/restart the networker services most of the time for a new client to
> get added properly now.
> This was not the behavior of Networker 6.0.1 or 5.5.
>
>
> -----Original Message-----
> From: Yura Pismerov [mailto:ypismerov AT TUCOWS DOT COM]
> Sent: Friday, July 11, 2003 12:57 PM
> To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
> Subject: Re: [Networker] Client Addition bug in 6.1.3
>
>
> Joel Fisher wrote:
>>
>> Sure has the symptoms of a DNS resolution problem... except that I can
>> resolve forward and reverse fine from both client and server.  Also
> why
>
>
> May I ask what is being used for the DNS resolution test ?
> Do you use nslookup or you do it with "getent hosts hostname/IP" ?
> I hope you are aware that for accurate test you should use either getent
> (supported on most platforms), or host (may be not supported
> everywhere), but not nslookup.
>
>
>> would restarting Networker alone make the "not a registered client"
>> error go away?  There is a bug in the process of creating a new
> client.
>> Sometimes it works fine, other times it requires a initial host entry,
>> other times I have to restart Networker.
>>
>> Joel
>>
>> -----Original Message-----
>> From: Yura Pismerov [mailto:ypismerov AT TUCOWS DOT COM]
>> Sent: Friday, July 11, 2003 1:13 PM
>> To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
>> Subject: Re: [Networker] Client Addition bug in 6.1.3
>>
>> So it sounds like DNS resolution problem on the server side.
>> Besides nsswitch.conf I'd double check /etc/resolv.conf to ensure that
>> all the nameservers listed there are valid.
>>
>> Joel Fisher wrote:
>>>
>>> We actually have 6.1.2, but I get this error quite frequently.  I've
>>> been able to "work around" it most of the time by adding a
> /etc/hosts
>>> entry for the client.  Then once the initial backup has run I can
>> remove
>>> that /etc/hosts entry.  Occasionaly I have to restart networker to
> get
>> a
>>> client added.
>>>
>>> One person on the list sent a suggestion on editing the
>>> /etc/nsswitch.conf file.  Check the archives.  I haven't tried it
> yet,
>>> but it seemed to work for him.
>>>
>>> Joel
>>>
>>> -----Original Message-----
>>> From: Donovan O'Brien [mailto:dobrien AT SIAC DOT COM]
>>> Sent: Friday, July 11, 2003 11:31 AM
>>> To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
>>> Subject: Re: [Networker] Client Addition bug in 6.1.3
>>>
>>> Most clients only have one IP address.  The others we add the alias
> to
>>> Networker (that brings up another bug with 6.1.3, if you have too
> many
>>> aliases and the nsr.res field spans more than one line, storage
> nodes
>>> fail
>>> to work, but that is another problem).
>>>
>>> Last night I added a client, its backup kicked off 6 hours later,
> that
>>> should be enough time for Networker to sync, and it failed.  This
>>> morning,
>>> I recycled the Networker processes then started another backup for
>> that
>>> same client which succeeded.
>>>
>
> --
> Note: To sign off this list, send a "signoff networker" command via
> email
> to listserv AT listmail.temple DOT edu or visit the list's Web site at
> http://listmail.temple.edu/archives/networker.html where you can
> also view and post messages to the list.
> =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
>
> --
> Note: To sign off this list, send a "signoff networker" command via email
> to listserv AT listmail.temple DOT edu or visit the list's Web site at
> http://listmail.temple.edu/archives/networker.html where you can
> also view and post messages to the list.
> =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=

--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=