How to configure time synchronization for two NTP servers

claudiols · ‎08-15-2013

We have IOSXR 4.2.1 on routers CRS3 and ASR9K with all recomended SMUs; we need to configure the time synchronization for two NTP servers with the configuration below, but the routers became unstable; synchronize with one NTP servers for some time, then switch to other NTP server, and keep doing this. Anyone know why this behavior?

ntp

authentication-key 1 md5 encrypted 01070F074F0A05

authenticate

trusted-key 1

server 10.192.32.32 prefer

server 10.192.32.33

source Loopback50

update-calendar

RP/0/RP0/CPU0:DFCRSDTC1#sh log | i ntp

Wed Jul 10 09:37:04.621 BRSPO

RP/0/RP0/CPU0:Jul 4 21:29:18 : ntpd[256]: %IP-IP_NTP-5-SYNC_LOSS : Synchronization lost : 10.192.32.32 : Peer unreachable or clock selection failed

RP/0/RP0/CPU0:Jul 4 21:29:18 : ntpd[256]: %IP-IP_NTP-5-HP_CONN_LOST : High priority NTP peer connection lost - Stratum 2->15.

RP/0/RP0/CPU0:Jul 4 21:29:18 : ntpd[256]: %IP-IP_NTP-5-ALL_CONN_LOST : All NTP peer connections failed.

RP/0/RP0/CPU0:Jul 4 21:29:27 : ntpd[256]: %IP-IP_NTP-5-HP_CONN_RECOVERED : High priority NTP peer connection recovered - Stratum 15->2.

RP/0/RP0/CPU0:Jul 4 21:30:21 : ntpd[256]: %IP-IP_NTP-5-SYNC_LOSS : Synchronization lost : 10.192.32.32 : Peer unreachable or clock selection failed

RP/0/RP0/CPU0:Jul 4 21:30:21 : ntpd[256]: %IP-IP_NTP-5-HP_CONN_LOST : High priority NTP peer connection lost - Stratum 2->15.

RP/0/RP0/CPU0:Jul 4 21:30:21 : ntpd[256]: %IP-IP_NTP-5-ALL_CONN_LOST : All NTP peer connections failed.

RP/0/RP0/CPU0:Jul 4 21:31:36 : ntpd[256]: %IP-IP_NTP-5-HP_CONN_RECOVERED : High priority NTP peer connection recovered - Stratum 15->2.

RP/0/RP0/CPU0:Jul 4 21:35:56 : ntpd[256]: %IP-IP_NTP-5-SYNC_LOSS : Synchronization lost : 10.192.32.33 : Peer unreachable or clock selection failed

RP/0/RP0/CPU0:Jul 4 21:35:56 : ntpd[256]: %IP-IP_NTP-5-HP_CONN_LOST : High priority NTP peer connection lost - Stratum 2->15.

RP/0/RP0/CPU0:Jul 4 21:35:56 : ntpd[256]: %IP-IP_NTP-5-ALL_CONN_LOST : All NTP peer connections failed.

RP/0/RP0/CPU0:Jul 4 21:40:11 : ntpd[256]: %IP-IP_NTP-5-HP_CONN_RECOVERED : High priority NTP peer connection recovered - Stratum 15->2.

RP/0/RP0/CPU0:Jul 4 21:50:52 : ntpd[256]: %IP-IP_NTP-5-SYNC_LOSS : Synchronization lost : 10.192.32.33 : System clock selection failed

RP/0/RP0/CPU0:Jul 4 21:50:52 : ntpd[256]: %IP-IP_NTP-5-HP_CONN_LOST : High priority NTP peer connection lost - Stratum 2->6.

RP/0/RP0/CPU0:Jul 4 21:59:26 : ntpd[256]: %IP-IP_NTP-5-HP_CONN_RECOVERED : High priority NTP peer connection recovered - Stratum 6->2.

RP/0/RP0/CPU0:Jul 4 22:25:07 : ntpd[256]: %IP-IP_NTP-5-SYNC_LOSS : Synchronization lost : 10.192.32.33 : System clock selection failed

RP/0/RP0/CPU0:Jul 4 22:25:07 : ntpd[256]: %IP-IP_NTP-5-HP_CONN_LOST : High priority NTP peer connection lost - Stratum 2->6.

RP/0/RP0/CPU0:Jul 4 22:56:16 : ntpd[256]: %IP-IP_NTP-5-SYNC_LOSS : Synchronization lost : 10.192.32.33 : Peer unreachable or clock selection failed

RP/0/RP0/CPU0:Jul 4 22:56:16 : ntpd[256]: %IP-IP_NTP-5-ALL_CONN_LOST : All NTP peer connections failed.

xthuijs · ‎08-15-2013

You might be peering with a server that doesnt like ntpv4 as much.

Try to configure version 3 under the ntp config and see if things stabalize.

also the show ntp trace might have some clues as to what the issue could be if that versioning is not the issue.

regards

xander

claudiols · ‎08-15-2013

Alexander, we tryied to use the ntpv3 but it do not worked. We have a TAC case and she thinks that could be CSCtd77162, but in this bug description the NTP server need to be with some inconsistency; we already checked on NTP servers administrators and there is not any inconsistency, both times are the same (these servers are used by many others Cisco and not Cisco routers without problems). I just open this forum to know if someone else have the same problem when configure two NTP servers.

xthuijs · ‎08-15-2013

Hi Claudio, that ddts is pretty generic to be honest but yes it is filed to address sync issues in the XR NTP algo.

The thing is that XR ntp clock selection is a bit different then iOS and follows the specs very closely which results in this erroneous loss behavior.

For instance, you could also see this issue with a sync loss if the update time is only 500msec off what it was before and that will result in a ntp sync loss rather then adjusting to it.

Also I wanted to mention that the ntp prefer is a bit of a misnomer in XR (since it follows the specs differently then IOS) and this knob was taken over from IOS really.

You might get some joy if you set it to one server only and see if that helps?

regards

xander