NTP (Network Time Protocol) is used to allow network devices to synchronize their clocks with a central source clock. For network devices like routers, switches or firewalls this is very important because we want to make sure that logging information and timestamps have the accurate time and date. If you ever have network issues or get hacked, you want to make sure you know exactly what and when it happened.
Normally a router or switch will run in NTP client mode which means that it will adjust its clock based on the time of a NTP server. Basically the NTP protocol describes the algorithm that the NTP clients use to synchronize their clocks with the NTP server and the packets that are used between them.
A good example of a NTP server is ntp.pool.org. This is a cluster of NTP servers that many servers and network devices use to synchronize their clocks.
NTP uses a concept called “stratum” that defines how many NTP hops away a device is from an authorative time source. For example, a device with stratum 1 is a very accurate device and might have an atomic clock attached to it. Another NTP server that is using this stratum 1 server to sync its own time would be a stratum 2 device because it’s one NTP hop further away from the source. When you configure multiple NTP servers, the client will prefer the NTP server with the lowest stratum value.
Cisco routers and switches can use 3 different NTP modes:
- NTP client mode.
- NTP server mode.
- NTP symmetric active mode.
The symmetric active mode is used between NTP devices to synchronize with each other, it’s used as a backup mechanism when they are unable to reach the (external) NTP server.
In the remaining of this tutorial I will demonstrate how to configure NTP on a Cisco router and switches.
This is the topology I will use:
The router on the top is called “CoreRouter” and its the edge of my network. It is connected to the Internet and will use one of the NTP servers from pool.ntp.org to synchronize its clock. The network also has two internal switches that require synchronized clocks. Both switches will become NTP clients of the CoreRouter, thus making the CoreRouter a NTP server.
First we will configure the CoreRouter on top. I will use pool.ntp.org as the external NTP server for this example. We need to make sure that the router is able to resolve hostnames:
CoreRouter(config)#ip name-server 184.108.40.206
I will use Google DNS for this. Our next step is to configure the NTP server:
CoreRouter(config)#ntp server pool.ntp.org
That was easy enough, just one command and we will synchronize our clock with the public server. We can verify our work like this:
CoreRouter#show ntp associations address ref clock st when poll reach delay offset disp ~220.127.116.11 .INIT. 16 - 64 0 0.000 0.000 16000. * sys.peer, # selected, + candidate, - outlyer, x falseticker, ~ configured
Above we see the show ntp associations command that tells us if our clock is synchronized or not. The ~ in front of the IP address tells us that we configured this server but we are not synchronized yet. You can see this because there is no * in front of the IP address and the “st” field (stratum) is currently 16.
There is one more command that gives us more information about the NTP configuration:
CoreRouter#show ntp status Clock is unsynchronized, stratum 16, no reference clock nominal freq is 250.0000 Hz, actual freq is 250.0000 Hz, precision is 2**24 reference time is 00000000.00000000 (00:00:00.000 UTC Mon Jan 1 1900) clock offset is 0.0000 msec, root delay is 0.00 msec root dispersion is 0.16 msec, peer dispersion is 0.00 msec loopfilter state is 'FSET' (Drift set from file), drift is 0.000000000 s/s system poll interval is 64, never updated.
The router tells us that we are unsynchronized and that there is no reference clock…we will just wait a couple of minutes and take a look at these commands again:
CoreRouter#show ntp associations address ref clock st when poll reach delay offset disp *~18.104.22.168 22.214.171.124 2 26 64 1 10.857 -5.595 7937.5 * sys.peer, # selected, + candidate, - outlyer, x falseticker, ~ configured
A few minutes later and the output has changed. The * in front of the IP address tells us that we have synchronized and the stratum is 2…that means that this NTP server is pretty close to a reliable time source. The “poll” field tells us that we will try to synchronize the time every 64 seconds. Let’s check the other command that we just saw:
CoreRouter#show ntp status Clock is synchronized, stratum 3, reference is 126.96.36.199 nominal freq is 250.0000 Hz, actual freq is 250.0000 Hz, precision is 2**24 reference time is D76513B4.66A4CDA6 (12:40:20.400 UTC Mon Jul 7 2014) clock offset is -5.5952 msec, root delay is 13.58 msec root dispersion is 7966.62 msec, peer dispersion is 7937.50 msec loopfilter state is 'CTRL' (Normal Controlled Loop), drift is -0.000000018 s/s system poll interval is 64, last update was 43 sec ago.
Our clock has been synchronized and our own stratum is 3, that makes sense since the public stratum server has a stratum of 2 and we are one “hop” away from it.
Cisco routers have two different clocks, they have a software clock and a hardware clock and they operate separately from each other. Here’s how to see both clocks:
CoreRouter#show clock 12:41:25.197 UTC Mon Jul 7 2014
CoreRouter#show calendar 12:43:24 UTC Mon Jul 7 2014
The show clock command shows me the software clock while the show calendar command gives me the hardware clock. The two clocks are not in sync so this is something we should fix, you can do it like this:
The ntp update-calendar command will update the hardware clock with the time of the software clock, here’s the result:
CoreRouter#show clock 12:42:31.853 UTC Mon Jul 7 2014
CoreRouter#show calendar 12:42:30 UTC Mon Jul 7 2014
That’s all I wanted to configure on the CoreRouter for now. We still have to configure two switches to synchronize their clocks.
The two switches will be configured to use the CoreRouter as the NTP server and I will also configure them to synchronize their clocks with each other. Let’s configure them to use the CoreRouter first:
SW1(config)#ntp server 192.168.123.3
Once again it might take a few minutes to synchronize but this is what you will see:
SW1#show ntp associations address ref clock st when poll reach delay offset disp *~192.168.123.3 188.8.131.52 3 21 64 1 2.5 1.02 15875. * master (synced), # master (unsynced), + selected, - candidate, ~ configured
SW1#show ntp status Clock is synchronized, stratum 4, reference is 192.168.123.3 nominal freq is 119.2092 Hz, actual freq is 119.2089 Hz, precision is 2**18 reference time is D765271D.D6021302 (14:03:09.835 UTC Mon Jul 7 2014) clock offset is 1.0229 msec, root delay is 14.31 msec root dispersion is 16036.00 msec, peer dispersion is 15875.02 msec
The clock of SW1 has been synchronized and its stratum is 4. This makes sense since it’s one “hop” further away from its NTP server (CoreRouter). Let’s do the same for SW2:
SW2(config)#ntp server 192.168.123.3
Let’s be patient for a few more minutes and this is what we’ll get:
SW2#show ntp associations address ref clock st when poll reach delay offset disp *~192.168.123.3 184.108.40.206 3 17 64 37 3.4 1.89 875.8 * master (synced), # master (unsynced), + selected, - candidate, ~ configured
SW2#show ntp status Clock is synchronized, stratum 4, reference is 192.168.123.3 nominal freq is 119.2092 Hz, actual freq is 119.2084 Hz, precision is 2**18 reference time is D765274D.D51A0546 (14:03:57.832 UTC Mon Jul 7 2014) clock offset is 1.8875 msec, root delay is 15.18 msec root dispersion is 1038.39 msec, peer dispersion is 875.84 msec
SW1 and SW2 are now using CoreRouter to synchronize their clocks. Let’s also configure them to use each other for synchronization. This is the symmetric active mode I mentioned before, basically the two switches will “help” each other to synchronize…this might be useful in case the CoreRouter fails some day:
SW1(config)#ntp peer 192.168.123.2
SW2(config)#ntp peer 192.168.123.1
After waiting a few minutes you’ll see that SW1 and SW2 have synchronized with each other:
SW1#show ntp associations address ref clock st when poll reach delay offset disp *~192.168.123.3 220.127.116.11 3 59 64 37 3.0 -0.74 877.4 +~192.168.123.2 192.168.123.3 4 50 128 376 2.2 -2.04 1.3 * master (synced), # master (unsynced), + selected, - candidate, ~ configured
SW2#show ntp associations address ref clock st when poll reach delay offset disp *~192.168.123.3 18.104.22.168 3 45 128 377 2.9 1.95 1.0 ~192.168.123.1 192.168.123.3 4 67 1024 376 1.8 2.40 1.4 * master (synced), # master (unsynced), + selected, - candidate, ~ configured
Great everything is now in sync.
Want to take a look for yourself? Here you will find the final configuration of each device.
hostname CoreRouter ! interface FastEthernet0/0 ip address 192.168.123.3 255.255.255.0 ! ip name-server 22.214.171.124 ! ntp server pool.ntp.org ntp update-calendar ! end
hostname SW1 ! interface FastEthernet0/24 ip address 192.168.123.1 255.255.255.0 ! ntp server 192.168.123.3 ntp peer 192.168.123.2 ! end
hostname SW2 ! interface FastEthernet0/24 ip address 192.168.123.2 255.255.255.0 ! ntp server 192.168.123.3 ntp peer 192.168.123.1 ! end
Are we done? Not quite yet…there are a few more things we can do with NTP. The CoreRouter and the two switches use unicast (UDP port 123) for synchronization but you can also use multicast or broadcast. Let me give you an example…
Multicast and Broadcast
If you have more than 20 network devices or a router that has limited system memory or CPU resources you might want to consider using NTP broadcast or multicast as it requires less resources. We can enable multicast or broadcast on the interface level.To demonstrate this I will add two routers below SW1 and SW2 that will synchronize themselves using multicast or broadcast. This is what it looks like:
I’ll configure SW1 to use multicast address 126.96.36.199 and SW2 will send NTP updates through broadcast: