dhcpd 4.1.1

[John Wobus]

We simply restart the server many times a day.
I assume a lot of sites do this though
many reduce restarts by adding hosts
using the programmatic interface.

By restarting a lot, we are less likely
to suffer from memory leak issues
and such, but more likely to suffer
from rare random startup and
"brief downtime" related problems.

We run redundant servers and restart the
two in series. I generally don't see issues
with the frequent startups thought it
undoubtedly slows response to clients.
As our site has grown, time until a restarted
server is responding has increased a bit.
I rarely notice anything more than that, but
recently did analyze a client's problem
that appeared to happen because
there was insufficient time between
the two servers' restarts to share
information about the client's new
lease.

John
Cornell University IT

P.S. I once proposed ISC make SIGHUP simply
'exec' itself with the same parms, in effect,
a restart with minimal downtime.  Otherwise,
downtime can be minimized by executing
(and re-executing) the daemon with 'system'
in a shell script loop.  That would fit
right in with a "nanny process" such as
is typical for the mysql daemon.

We don't do that: we have code that polls
the old process very often to determine
when to start the new instance.

On Jul 19, 2013, at 10:19 AM, Steve Clark wrote:

> On 07/19/2013 10:12 AM, Jason Brandt wrote:
>> What is your need for reloading the config file?  Adding networks  
>> and such, always requires a reload, but there are ways to add  
>> hosts, subclasses, etc that don't require reloading the config.
>>
>>
> It would be for adding new networks.
>
> Thanks,
> Steve
>
>
>> On Fri, Jul 19, 2013 at 9:06 AM, Steve Clark <sclark at netwolves.com>  
>> wrote:
>> On 07/19/2013 09:31 AM, Steven Carr wrote:
>>> Man page for DHCPD 4.1.1-P1 states:
>>>
>>> Note: We get a lot of complaints about this.   We realize that it  
>>> would be nice if one could send a SIGHUP to the server and have it  
>>> reload
>>> the database.   This is not technically impossible, but it would  
>>> require a great deal of work, our resources are extremely limited,  
>>> and they can be better spent elsewhere.   So please don't complain  
>>> about this on the mailing list unless you're prepared to fund a  
>>> project to implement this feature, or prepared to do it yourself.
>>>
>>> ;)
>>>
>>>
>> Hi Steven,
>>
>> Sorry for not reading the man page more carefully, I missed that  
>> part and went to end
>> where most man pages talk about using SIGHUP to cause the process  
>> to re-read its
>> conf file.
>>
>> Anyway thanks for the pointer.
>>
>>
>>> On 19 July 2013 14:12, Steve Clark <sclark at netwolves.com> wrote:
>>> On 07/19/2013 08:32 AM, Niall O'Reilly wrote:
>>>> On 19 Jul 2013, at 13:27, Steve Clark wrote:
>>>>
>>>>> Is there a way to make dhcpd re-read it conf file. I did
>>>>> looked in the man page and did a quick google but didn't
>>>>> see a way. Maybe via omapi?
>>>> 	Restarting dhcpd works for me.
>>>>
>>>> 	I suspect that the kind of method you suggest would
>>>> 	probably require stalling the service for a similar
>>>> 	duration as that caused by a restart.
>>>>
>>>> 	Niall O'Reilly
>>>>
>>> I would really like something like send a SIGHUP, etc.
>>>
>>>
>>
>
> -- 
> Stephen Clark
> NetWolves
> Director of Technology
> Phone: 813-579-3200
> Fax: 813-882-0209
> Email: steve.clark at netwolves.com
> http://www.netwolves.com
> _______________________________________________
> dhcp-users mailing list
> dhcp-users at lists.isc.org
> https://lists.isc.org/mailman/listinfo/dhcp-users