failover timeout

[Glenn Satchell]

The theory was that any logic required to determine if the partner was 
truly down was all external to dhcpd and often very specific to the 
site. As there is an existing way to remotely control dhcpd using 
omshell, the idea was to leave the determination to the site admin.

The auto-partner-feature in version 4.2.0 is probably does what you 
want. Maybe download the source and have a look at the new documentation.

regards,
-glenn

On 04/12/10 20:36, Abu Abdulla alhanbali wrote:
> Thanks Glenn. i was thinking the server will take a decision by itself
> (according to some logic) to take the full traffic.
>
> On Fri, Apr 9, 2010 at 8:29 AM, Glenn Satchell
> <glenn.satchell at uniq.com.au>  wrote:
>> 200k leases with one week duration. One week is 604800 seconds, so 200k
>> leases renewing weekly would be about one every 3 seconds. Even if clients
>> renew more often (typically 50% of lease time, or customers reboot CPE) this
>> is still a very low number that would not require a very powerful computer.
>> Given the high number of clients, failover or HA of some sort would be a
>> good recommendation.
>>
>> regards,
>> -glenn
>>
>> On 04/09/10 00:39, Glenn Satchell wrote:
>>>
>>> 4,2.0 (currently alpha) has a configurable auto-partner-down feature.
>>> For all earlier releases the server stays in communications-interrupted
>>> and *never* goes to partner down mode without manual action.
>>> Partner-down is the mode where the surviving server can allocate
>>> addresses previously given out by the failed server.
>>>
>>> Having said that it is pretty easy to write a script that monitors the
>>> other dhcp server, and uses omshell to switch to partner down when it
>>> detects that the other server has gone away. When the failed server
>>> comes back it will come out of partner-down mode and return to normal as
>>> part of the startup negotiations.
>>>
>>> There was a post to the list by David Hankins on 25 Feb this year
>>> mentioning this which gives a lot more detail.
>>>
>>> regards,
>>> -glenn
>>>
>>> On 04/08/10 13:48, Abu Abdulla alhanbali wrote:
>>>>
>>>> this is not what i mean. if the primary servers is down, how long will
>>>> it take the secondary server to start assigning the IPs that are
>>>> reserved for the primary server. you can do this manually using the
>>>> omshell but i don't know if this will be done automatically or not.
>>>>
>>>> On Wed, Apr 7, 2010 at 7:55 PM, Banana Flex<flex.banana at bluewin.ch>
>>>> wrote:
>>>>>
>>>>> hi
>>>>>
>>>>> you can do a collaboration between two servers that send the same
>>>>> subnet, see the man page of dhcpd.conf for many information. you can
>>>>> also configure the timeout. 3 seconds is a good choice
>>>>>
>>>>> have fun
>>>>> Banana
>>>>>
>>>>>
>>>>> On Apr 7, 2010, at 5:41 AM, Abu Abdulla alhanbali wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> i would like to know how long it will take the secondary DHCP server
>>>>>> to start assigning IPs for all users (including the ranges dedicated
>>>>>> to the primary) in case the primary DHCP is down? is it a
>>>>>> configurable?
>>>>>>
>>>>>> i would like also to ask if the DHCP can handle 200K users or above
>>>>>> assuming the H/W is capable and the lease time is big enough (more
>>>>>> than one week). is it recommended to run multiple instances or let one
>>>>>> instance handle this number.
>>>>>>
>>>>>> thanks