[Pacemaker] Problem with gratuitous arps in IPaddr2

Andrew Beekhof andrew at beekhof.net
Wed Oct 21 10:11:52 EDT 2009


Fixed:
  http://hg.linux-ha.org/agents/rev/36323bc53465

On Sun, Oct 11, 2009 at 10:49 PM, Andrew Beekhof <andrew at beekhof.net> wrote:
> On Wed, Sep 23, 2009 at 4:28 PM, Lars Ellenberg
> <lars.ellenberg at linbit.com> wrote:
>> On Wed, Sep 23, 2009 at 03:05:32PM +0200, Christoph Mitasch wrote:
>>> Hello,
>>>
>>> I just read the following thread:
>>> http://www.gossamer-threads.com/lists/linuxha/users/56429?do=post_view_threaded#56429
>>>
>>> What does this mean for the moment?
>>> Is the IPaddr2 resource not working correctly at the moment on CentOS 5?
>>>
>>> What is the way to go at the moment?
>>
>> Dunno.
>> but one could replace the call to /usr/lib/heartbeat/send_arp
>> in IPaddr2 with calls to /usr/bin/arping,
>> where arping come from the package called (in debian) iputils-arping.
>>
>> heartbeat send_arp:
>> send_arp -i $interval_ms -r $repeat_count -p $some_pid_file $DEVICE $IP ${MAC:-auto} not_used not_used
>>
>> iputils-arping:
>> (
>> arping -U -b -c $repeat_count -w $repeat_count -I $DEVICE -s $IP $IP &
>> arping -A -b -c $repeat_count -w $repeat_count -I $DEVICE -s $IP $IP &
>
> Is this second one (with -A) really necessary?
> My arping reading of the code indicates it shouldn't be.
>
>> wait
>> )
>>
>> should be almost functionally equivalent...
>
> It should be more than that :-)
> send_arp.linux.c is the arping code with a few changes to support the
> interface exposed by the old send_arp binary.
>




More information about the Pacemaker mailing list