Bacula-users

Re: [Bacula-users] timeout contacting storage daemon on system with multiple interfaces (SOLVED)

2008-06-04 16:13:24
Subject: Re: [Bacula-users] timeout contacting storage daemon on system with multiple interfaces (SOLVED)
From: Martin Simmons <martin AT lispworks DOT com>
To: bacula-users AT lists.sourceforge DOT net
Date: Wed, 4 Jun 2008 21:13:11 +0100
>>>>> On Wed, 04 Jun 2008 14:16:33 -0400, Josh Fisher said:
> 
> Martin Simmons wrote:
> >>>>>> On Wed, 4 Jun 2008 16:21:27 +0200, Ariano Bertacca said:
> >>>>>>             
> >>
> >> I forgot to mention, that i don't think this is the perfect solution.  
> >> This works, but i have to tweak the clients resolver to get bacula to  
> >> use the right ip. Since the server has multiple interfaces and  
> >> multiple names for these interfaces, i would prefer to configure the  
> >> name/ip i want to use inside bacula instead of relying on the fqdn.
> >>     
> >
> > The client connects using the Address field of the Director's Storage
> > resource, so it doesn't have to be the fqdn, just something that resolves to
> > the correct IP address on all machines.
> >
> > What kind of syntax would you use in the Director's config to specify a
> > different name/ip for different clients?  Maybe it should allow a mapping 
> > from
> > the client's subnet to the SD address?
> >
> >   
> 
> I believe the problem is when SD is multi-homed. For example, the 
> clients on the 10.1.2.0/24 network connect to the SD at 10.1.2.10 and 
> the clients on 10.1.3.0/24 connect to the SD at 10.1.3.10.

Yes, that is one case.  OTOH, you could make it more general, so that clients
on the x.x.x.x/X network should connect to a.a.a.a and those on the y.y.y.y/Y
should connect to b.b.b.b.  I.e. the IP for the SD might not be within the
client's own subnet, but it might still vary.


> Unfortunately, it is not possible to define two Storage resources in the 
> Director that point to the same Device resource in the SD, and the SD's 
> Address directive cannot take more than one parameter. So it is not 
> possible to define this in Bacula, and the only choice is to use a FQDN.

Right.  Even if there could be more than one Address directive, the director
would not know which one to pass to the client.


> Rather than tweaking the resolver on the client, however, it would be 
> better to put two A records for the FQDN in DNS. The FD would then have 
> to pull all IPs from its DNS query and favor an IP in its own subnet 
> over the others. I doubt the FD currently does this (though it might be 
> worth a try). It probably pulls only the first IP, in which case the IP 
> chosen would alternate between the two because the resolver normally 
> load-balances in round-robin fashion by changing the order of the 
> answers. So this likely wouldn't work without a change to the FD to 
> cause it to parse the addresses returned from the resolver and look for 
> an IP in its own subnet, rather than just choosing the first answer.

Deliberately putting unroutable IP addresses in the DNS sounds wrong to me,
because it will break communication with the SD for all other programs that
don't check both addresses.

__Martin

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

<Prev in Thread] Current Thread [Next in Thread>