full socket buffers

[Mirek Lauš]

On Fri, Feb 11, 2011 at 1:18 PM, Chris Marget <chris at marget.com> wrote:
> On Fri, Feb 11, 2011 at 5:40 AM, Mirek Lauš <mirek at admino.cz> wrote:
>> On Wed, Feb 9, 2011 at 6:25 PM, Friesen, Don SSBC:EX
>> <Don.Friesen at gov.bc.ca> wrote:
>>>>>
>>>>> No. It is completely random. When the issue occur, the netstat -s -p
>>>>> udp show continual
>>>>> loss of UDP packets due to full socket buffers.
>>>>>
>>>>> I already stated that the kern.ipc.maxsockbuf is set to 8MB and
>>>>> net.inet.udp.recvspace is set to 1MB. How can I monitor buffer usage?
>>>>
>>>>And no, there is no heavy network traffic at the time.
>>>>
>>>
>>>  What is the buffer full of?  What sort of packets are coming off?  Can you capture a snoop and analyze the ports
>>
>> How do I find what is the buffer full of? There is only DHCP and LDAP
>> syncrepl traffic on the server.
>
> These are per-socket buffers.  They'll be full of whatever UDP
> datagrams were addressed to the port in question.  (you're losing
> incoming datagrams, right?)

Yes. Loosing incoming UDP datagrams which causes DHCPD process to starve.
But restarting DHCPD process solves the situation ...

Regards,
Miroslav