DHCP-server failover feature-request

   *** From dhcp-server -- To unsubscribe, see the end of
this message. ***


>Subject: Re: DHCP-server failover feature-request
>From: johan Henęs <johanhenes.no>
>To: dhcp-serverisc.org
>Date: Sat, 18 Feb 2006 17:44:28 +0100
>
>On Sat, 2006-02-18 at 13:50 +1100, Glenn Satchell wrote:
>> It's not manual - it works perfectly well without
touching it. It you read 
the 
>> dhcpd.conf man page, setting the server into
partner down mode is only 
required 
>> for "extended" outages, where extended
means something like greater than half 
>> the lease time.
>> 
>Ah - sorry for me not beeing accurate enough - Yes it
works perfectly,
>BUT server outages caused by disk-crashes etc. tend to
often last longer
>than half the lease time... 

Hmm, maybe you need longer lease time, or RAID 1 disks on
your servers?
Disks are so low in cost these days that I build every
server with
mirrored disks. Even my workstation has mirrored disks after
the last
hardware upgrade.

>> There were some scripts sent to the list a few
months back that ran on
>> each server and monitored the peer. If the peer was
uncontactable for
>> some amount of time then it used OMAPI to switch
the local server to
>> partner down.
>> 
>That is what i need - i will try to find them.. 

This is the thread I was referring to. I've been looking at
this with a
view to implementing it myself on a few different dhcp
networks.

http://marc.theaimsg
roup.com/?l=dhcp-server&w=2&r=1&s=Splitting+of+I
P+address+on
+DHCP+failover&q=b

>> If you have, say, 8 hour leases, then a 30 minute
outage should be
>> acceptable. If you keep switching to partner down
for short outages
>> then the serves have a lot of work to do to
resynchronise lease files.
>
>I do totally agree - Maybe a
"Auto-partner-down-time='x mins'" would be
>a better config-parameter ??
>I would appreciate it anyway 

I think that would be a more useful option, but it is
achievable with
things external to the dhcp server process. Kind of like the
unix
philosophy of having many small tools, rather than one huge
one that
tries to do everything well.

regards,
-glenn


------------------------------------------------------------
-----------
List Archives : http://www.isc.org/ops/
lists/
Unsubscribe   : http://www.
isc.org/sw/dhcp/dhcp-lists.php    
-or-          : mailto:dhcp-server-requestisc.org?Subject=unsubscribe  
------------------------------------------------------------
-----------

   *** From dhcp-server -- To unsubscribe, see the end of
this message. ***

On Sun, 2006-02-19 at 22:17 +1100, Glenn Satchell wrote:

> This is the thread I was referring to. I've been
looking at this with a
> view to implementing it myself on a few different dhcp
networks.
> 
> http://marc.theaimsg
roup.com/?l=dhcp-server&w=2&r=1&s=Splitting+of+I
P+address+on
> +DHCP+failover&q=b
> 

Thanks a lot for Your help, Glenn !!
This is just what i need 

> I think that would be a more useful option, but it is
achievable with
> things external to the dhcp server process. Kind of
like the unix
> philosophy of having many small tools, rather than one
huge one that
> tries to do everything well.

I do agree with You, but I do think that there are
"customers" out there
expecting a failover-solution to be fully automatic. Even if
this is
achievable as You say by other tools, my impression is that
this is one
of the reasons why company's often buy solutions from M$
(god forbid),
rather than hiring expensive people with the knowledge on
how to do
things right. 
(As I told in my first mail in this thread, I haven't been
working with
this kind of stuff in a while, but the customer contacted me
because he
had gotten an offer for two W2k3 servers in a cluster (!)
from another
company, as a replacement to my old setup  

- Again - thanks a lot for Your help !!

Regards,

Johan


------------------------------------------------------------
-----------
List Archives : http://www.isc.org/ops/
lists/
Unsubscribe   : http://www.
isc.org/sw/dhcp/dhcp-lists.php    
-or-          : mailto:dhcp-server-requestisc.org?Subject=unsubscribe  
------------------------------------------------------------
-----------