[ntp:questions] Re: Servers just doen't work (after following the troubleshooting page)

Per Hedeland per at hedeland.org
Wed Sep 28 20:44:59 UTC 2005


In article <Pine.LNX.4.62.0509281201520.10217 at localhost.localdomain> S P
Arif Sahari Wibowo <arifsaha at yahoo.com> writes:
>On Wed, 28 Sep 2005, Per Hedeland wrote:
>> I.e. if the address in question *is* bound to a socket, the 
>> bug will not occur - and the OP claimed that ntpd *did* have 
>> all the IP addresses bound (but I suspect he may be mistaken).
>
>You are correct. I misunderstood you. I thought you meant that 
>each address is not listed more than once. In fact, the 
>secondary IPs was not listed at all. Giving -L to the ntpd 
>indeed make it listen to all IPs. So indeed it may work with my 
>original setup?

Yes. Actually I'm running that very version of ntpd in a commercial
product, using -L to get it to serve time on a "virtual" IP address for
a purpose very similar to what you wanted. But while it makes sense in
that particular context, I'm not sure it does in a general NTP
server/service setup (the product in question has nothing to do with
NTP, it's just an "internal" function there).

>Is there any cost associated on listening to secondary / virtual 
>IPs?

Not significant in "normal" cases I'd say, and whether they're secondary
/ virtual or not is irrelevant. I think part of the "controversy" around
this arose due to people running ntpd on web servers that had a
gazillion "virtual" IP addresses for "virtual hosting", and having no
interest in having ntpd listen on all of those. Watching for packets on
a gazillion sockets does have performance and general resource
consumption significance, in fact on some platforms ntpd might get into
serious trouble by trying.

--Per Hedeland
per at hedeland.org




More information about the questions mailing list