|
|
|
|
I have a pair of DHCP servers that specify 2 NTP servers for each pool. The pool definitions are exactly the same on each peer, and they are set to split 128. We have recently set up 2 additional NTP servers. I know the order of the NTP servers listed is in order of preference. Is there a recommended practice in order for load balancing the clients usage of the NTP servers? All of the NTP servers are the same strata. Does it make sense to define different NTP server options on each DHCP server? For example on DHCP1 use NTP1 and NTP2 and on DHCP2 use NTP3 and NTP4. Or does it make more sense to set the NTP servers in a round-robin fashion in pool definitions? For example pool1 defines NTP1, NTP2, NTP3, NTP4, then pool2 defines NTP2, NTP3, NTP4, NTP1, etc...
_______________________________________________
dhcp-users mailing list
[hidden email]
https://lists.isc.org/mailman/listinfo/dhcp-users
|
|
|
No need. The NTP client will query all available NTP services and choose one to follow. This query/choose continues as long as the client is running, so it
knows it is getting the best time source available. As the time service stabilize, the NTP client will slowly increase the time period between queries, to reduce the load on the NTP servers. If there is a disruption, the client will simply reset to a short
time period, and when thing settle down again, it will again, slowly increase the time period between queries.
Run ‘ntpq –p’ on your NTP client. The ‘poll’ column is the time period between polling (On my systems, it is 1024 seconds). The ‘when’ column is the number
of seconds since the last poll of the system in the ‘remote’ column. The asterisk in the first column, signifies the selected server. A plus signifies a selected server. A minus signifies a discarded server (good, but outside the balance of the selected
servers) and a space signifies a system discarded for high stratum level.
From: [hidden email] [mailto:[hidden email]]
On Behalf Of [hidden email]
Sent: Tuesday, February 16, 2016 9:41 AM
To: [hidden email]
Subject: Different NTP options on peers
I have a pair of DHCP servers that specify 2 NTP servers for each pool. The pool definitions are exactly the same on each peer, and they are set to split 128. We have recently set up 2 additional NTP servers. I know the order of the NTP
servers listed is in order of preference. Is there a recommended practice in order for load balancing the clients usage of the NTP servers? All of the NTP servers are the same strata. Does it make sense to define different NTP server options on each DHCP server?
For example on DHCP1 use NTP1 and NTP2 and on DHCP2 use NTP3 and NTP4. Or does it make more sense to set the NTP servers in a round-robin fashion in pool definitions? For example pool1 defines NTP1, NTP2, NTP3, NTP4, then pool2 defines NTP2, NTP3, NTP4, NTP1,
etc...
_______________________________________________
dhcp-users mailing list
[hidden email]
https://lists.isc.org/mailman/listinfo/dhcp-users
|
|
|
In reply to this post by matt.southward@gmail.com
On 16-Feb-16 11:15, [hidden email] wrote:
> Message: 5
> Date: Tue, 16 Feb 2016 16:15:36 +0000
> From: "Stier, Matthew" < [hidden email]>
> To: Users of ISC DHCP < [hidden email]>
> Subject: RE: Different NTP options on peers
> Message-ID:
> < [hidden email]>
> Content-Type: text/plain; charset="utf-8"
>
> No need. The NTP client will query all available NTP services and choose one to follow. This query/choose continues as long as the client is running, so it knows it is getting the best time source available. As the time service stabilize, the NTP client will slowly increase the time period between queries, to reduce the load on the NTP servers. If there is a disruption, the client will simply reset to a short time period, and when thing settle down again, it will again, slowly increase the time period between queries.
This is true if your clients are NTP. If they're Windoze clients (using
win32 time) they just ask the first one.
They don't do any of the NTP sync algorithms. They just ask, use what
they get, and some long time later, ask again.
There are lots of more accurate ways to manage time. But windows
doesn't care - as long as you can sync with your kerberos tokens, it's
happy. That needs maybe 5 mins accuracy; win32time can't do better than
a few seconds. If you really care about time (or you want your
makefiles/source control repos to run on a shared disk in an active
development environment), you can get a real NTP client for windows, but
almost no one does.
SNTP is midway between the two.
Timothe Litt
ACM Distinguished Engineer
--------------------------
This communication may not represent the ACM or my employer's views,
if any, on the matters discussed.
_______________________________________________
dhcp-users mailing list
[hidden email]
https://lists.isc.org/mailman/listinfo/dhcp-users
|
|
|
In reply to this post by matt.southward@gmail.com
[hidden email] wrote:
> Does it make sense to define different NTP server options on each DHCP server? For example on DHCP1 use NTP1 and NTP2 and on DHCP2 use NTP3 and NTP4. Or does it make more sense to set the NTP servers in a round-robin fashion in pool definitions? For example pool1 defines NTP1, NTP2, NTP3, NTP4, then pool2 defines NTP2, NTP3, NTP4, NTP1, etc...
It probably makes little difference.
If your intention is to spread the load, then don't list all servers in each instance - as pointed out, NTP will query all of them periodically. But NTP doesn't really load a server up much at all.
TBH, I run NTP on a couple of machines, and I never notice the load it creates - apart from when they get abused by some scumbag doing a DDoS on someone.
_______________________________________________
dhcp-users mailing list
[hidden email]
https://lists.isc.org/mailman/listinfo/dhcp-users
|
|
|
On 2/17/16 02:15, Simon Hobson wrote:
> [hidden email] wrote:
>
>> Does it make sense to define different NTP server options on each DHCP server? For example on DHCP1 use NTP1 and NTP2 and on DHCP2 use NTP3 and NTP4. Or does it make more sense to set the NTP servers in a round-robin fashion in pool definitions? For example pool1 defines NTP1, NTP2, NTP3, NTP4, then pool2 defines NTP2, NTP3, NTP4, NTP1, etc...
>
> It probably makes little difference.
> If your intention is to spread the load, then don't list all servers in each instance - as pointed out, NTP will query all of them periodically. But NTP doesn't really load a server up much at all.
> TBH, I run NTP on a couple of machines, and I never notice the load it creates - apart from when they get abused by some scumbag doing a DDoS on someone.
>
Everyone should check to be sure you have "disable monitor" set in your ntp.conf file or else
you contribute to the DDoS attacks coming from NTP services around the world.
--
Dave Calafrancesco
_______________________________________________
dhcp-users mailing list
[hidden email]
https://lists.isc.org/mailman/listinfo/dhcp-users
|
|
Locked
smime.p7s (6K)