High Availability NTP configuration issue

[KNSITReseau]

Hi,

We encounter some issues for setting up a NTP high availability architecture.

We have 2 GPS appliance which provide stratum 2 NTP services and a 3750x which redistribute NTP to all Vlans (approx. 100).

The issue is when the 3750x reboot it broadcast bad time values (2 march 1993) until it recover his correct synchronization with the GPS Appliance.

We have no alternative for the architecture because NTP clients (Industrial equipments) couldn't have a default gateway requires a IP address in each vlan to broadcast the NTP service.

Thanks,

Jérémy 

[Philip D'Ath]

Make sure you have not got the line "ntp master" in your 3750 configuration.

[KNSITReseau]

The configuration of the 3750x is :

ntp logging

ntp server X.X.X.X prefer

ntp server X.X.X.X

Nothing more.

Thanks

PS : IOS ver : 15.0(2)SE9

[johnd2310]

Hi,

I think the problem is your switches do not have a real-time clock and will always have the wrong time until they sync with NTP. 

Why not put a linux server on your network with an interface in each vlan and let it be the NTP for the industrial equipment. The linux can sync from the appliances.

Thanks

John

[KNSITReseau]

Hello,

I've solved my issue by using an event manager applet on syslog events.

Thanks

[Leo Laohoo]

The issue is when the 3750x reboot it broadcast bad time values (2 march 1993) until it recover his correct synchronization with the GPS Appliance.

Depends on the IOS the 3750X is running this can take between 5 & 13 minutes.  Looking at the network diagram, it tells me the two NTP boxes are connected to a single switch/stack, however, the network below the 3750X isn't very clear.  

Is this saying that the 3750X is the core?

Make sure you have not got the line "ntp master" in your 3750 configuration.

I agree with this recommendation.

Another question I would like to throw back is:  What kind of application that is very delicate to NTP?  I have several site "clocks" which require access to NTP, however, they aren't that sensitive to the switch rebooting.

[KNSITReseau]

The 3750x is the core of the NTP architecture, it just provide NTP services for all vlans.

Other informations : NTP clients (industrial equipment) request for NTP every 500ms (time is very important for logs and process report), when the 3750x reboot, clients take the bad time value.

 There is no way for blocking the NTP service on the switch until it's synchronized with the GPS appliance ?

[Leo Laohoo]

NTP clients (industrial equipment) request for NTP every 500ms (time is very important for logs and process report), when the 3750x reboot, clients take the bad time value.

Wow.  I've never seen this kind of chatty application before.  

Is the 3750X the core of the a flat Layer 2 network? 

[KNSITReseau]

The 3750x has one IP address in each vlans, we don't need routing protocols between the switch and the clients.

NTP clients are configured with the switch IP in correct vlan.

[Leo Laohoo]

Does this mean every clients connect to the 3750X?  And there are no other switch in the network?

[KNSITReseau]

No there is many other switches between the NTP clients and server.

[Leo Laohoo]

So get the NTP clients synch with their immediate switch.  The immediate switch then synch with their distro switch.  Distro switch synch with the core switch.   Core switch synch to the NTP server.