HP ProCurve 2910 al - SNTP difference

[ChimChim123]

Hi,

is it possible to configure SNTP on the ProCurve Switch to sync the time before there is a difference of 4 seconds??

This is the log message I get a few days a week:

SNTP: Updated time by -4 seconds

But 4 seconds are too much! 

Thx

-Chris

[Matcol]

The switch won't know the time divergence has reached 4 seconds until it tries to sync time, so the best approach is to reduce the sntp poll-interval.

Alternatively, you might have multiple NTP sources that diverge from each other?

[ChimChim123]

Yes, that's what I expect from NTP. The poll intervall on my switch is set to 300 seconds and I have 2 NTP Servers configured:

timesync sntp
sntp unicast
sntp 300
sntp server priority 1 192.168.70.53 3
sntp server priority 2 192.168.70.21 3

Our 2 NTP Servers have the same time, I also tried to change the NTP Version on the Switch, but same results.

I can even see the NTP Packets on my firewall, so the ProCurve Switch really poll every 300 seconds, but only if the 4 seconds are reached he updates itself. This happens regulary every 2 or 3 days as I can see in the logfiles:

Mar 11 23:48:38 10.0.0.6 00413 SNTP: Updated time by -4 seconds
Mar 8 02:18:46 10.0.0.6 00413 SNTP: Updated time by -4 seconds
Mar 10 01:03:42 10.0.0.6 00413 SNTP: Updated time by -4 seconds
Mar 19 18:38:22 10.0.0.6 00413 SNTP: Updated time by -4 seconds
Mar 17 19:38:26 10.0.0.6 00413 SNTP: Updated time by -4 seconds
Mar 15 20:58:30 10.0.0.6 00413 SNTP: Updated time by -4 seconds
Mar 13 19:58:34 10.0.0.6 00413 SNTP: Updated time by -4 seconds

So the question is why he polls every 300 seconds but nothing happens?

[Univ_of_Twente]

I found this discussion when I was searching for the exact same behavior: our ProCurve switches (various models, we currently have 2600/2610/2615/2620/2824/2915/2920/3800/5300/5400 in our network) have the same issue.

It never really bothered us until we had to debug a few issues where these clock inaccuracies were very had to compensate for, so we opened a support case.

As it turns out, this is by design in the SNTP implementation on the switches: the switch polls at the configured interval, but it will only correct the internal clock when it deviates more than a certain absolute threshold from the time on the (S)NTP server. That threshold seems to be 4 seconds. So ProCurve SNTP really is a simple implementation and not a full NTP client.

HP have just released x.16.01.yyyy software for various platforms, and among other enhancements, it includes a full NTP client implementation. This NTP client should provide perfect sync with NTP servers; I just installed it on a first switch to give it a try and after configuring NTP it already shows a more precise correction:

Feb 9 12:34:06 testswitch 04908 ntp: The system clock time was changed by -1 sec 196580675 nsec.

I'm afraid this doesn't help you with the 2910al since W.15.14.xxxx is the last maintained branch for your platform, but if you also have more current switches in your network, you may want to consider upgrading (especially after x.16.01.yyyy has matured a bit more).