[questions] Re: ntpd doesn't follow minpoll/maxpoll interval

[Sameera Unnisa]

Hi ,

I have my ntp.conf as below

"pool -4 server1 true minpoll 6 maxpoll 10 iburst"
"pool -6 server1 true minpoll 6 maxpoll 10 iburst"

I could see after 3rd request the very next 4th request is happening
earlier than minpoll interval i.e 2^6 sec.


Is there a way to give proper command to follow min and max poll interval?

Kindly help

Regards
Sameera

Hi ,
I have my ntp.conf as below
"pool -4 server1 true minpoll 6 maxpoll 10 iburst"
"pool -6 server1 true minpoll 6 maxpoll 10 iburst"
I could see after 3rd request the very next 4th request is happening
earlier than minpoll interval i.e 2^6 sec.
Is there a way to give proper command to follow min and max poll interval?
Kindly help
Regards
Sameera

[Harlan Stenn]

Sameera,

What version of ntpd are you running?

Why are you using the 'true' keyword?

Do you understand what the iburst keyword does?

What is server1, an NTP server or a source of NTP pool names?

Why are you querying server1 for IPv4 and IPv6 separately?

Thanks...

H

Post by Sameera Unnisa
Hi ,
I have my ntp.conf as below
"pool -4 server1 true minpoll 6 maxpoll 10 iburst"
"pool -6 server1 true minpoll 6 maxpoll 10 iburst"
I could see after 3rd request the very next 4th request is happening
earlier than minpoll interval i.e 2^6 sec.
Is there a way to give proper command to follow min and max poll interval?
Kindly help
Regards
Sameera
Hi ,
I have my ntp.conf as below
"pool -4 server1 true minpoll 6 maxpoll 10 iburst"
"pool -6 server1 true minpoll 6 maxpoll 10 iburst"
I could see after 3rd request the very next 4th request is happening
earlier than minpoll interval i.e 2^6 sec.
Is there a way to give proper command to follow min and max poll interval?
Kindly help
Regards
Sameera

[Sameera Unnisa]

What version of ntpd are you running?

4.2.8p15

Why are you using the 'true' keyword?

Its been there in my conf file by default was it not recommended?

Do you understand what the iburst keyword does?

Correct me if i'm wrong , it sends the burst of packet at time for early

synchronisation

What is server1, an NTP server or a source of NTP pool names?

An NTP server

Why are you querying server1 for IPv4 and IPv6 separately?

There is a limitation that I am getting first IPv4 and then IPv6 so

using it separately

Thanks

[Dave Hart]

Post by Sameera Unnisa
I have my ntp.conf as below
"pool -4 server1 true minpoll 6 maxpoll 10 iburst"
"pool -6 server1 true minpoll 6 maxpoll 10 iburst"
I could see after 3rd request the very next 4th request is happening
earlier than minpoll interval i.e 2^6 sec.

I see two distinct issues that may explain what you're seeing, but note
it's not entirely clear what you're seeing. You don't mention how you're
seeing the requests. Are you looking at a packet capture? Does the early
request occur on both IPv4 and IPv6?

1. The "pool" directive, as opposed to "server", instructs ntpd to attempt
to spin up as many associations for each hostname as possible, up until
ntpd has maxclock associations. With the above two lines being the
entirety of ntp.conf, I'd expect "tos maxclock 10" to be the default. If
server1 resolves to more than one A (IPv4) or AAAA (IPv6) record, ntpd will
solicit each in turn until 8 more associations are started. The "POOL"
associations, as seen in `ntpq -p` output, count for 2 of the 10. At the
risk of wandering into the weeds a bit, the initial solicitation request
appears just like a normal NTP client request, but the response is not
processed for time. Instead, it results in starting an association with
the responding IP address and scheduling the next solicitation, if any.

2. "iburst" instructs ntpd to initially send a sequence of 6 requests at
least 2 seconds apart, assuming a response is received for each, to speed
initial synchronization. If the association has become unreachable for
some time, iburst is activated again when responses are once again
received. Wandering in weeds again, stay away from "burst", which is
intended for synchronization occurring over an intermittent connection,
such over a dialup modem connection, where very high minpoll is used to
keep down costs by polling as infrequently as every 36 hours. It causes
the burst behavior on every query, rather than only initially.

If you continue to have problems, consider providing the output of `ntpq
-crv -w -clpeers` redacted of any names or addresses you wish.

Cheers,
Dave Hart