Strange problem with NTP on Catalyst 3550

[m.brausen]

Hi,

I am puzzled by a somewhat strange problem concerning NTP and three clusters of two Catalyst 3550-48 SMI: When we put them in use, I configured NTP on all switches which seemed to be running fine. Lately I noticed those three clusters would not longer sync with my NTP server (Linux box, xntpd 4.2.0). All other 3550 do fine (difference: they are not stacked).

What I notice is, that the server is still configured but offset and dispersion are very high so that they don't like my server any longer - but why? All switches are within the same network segment.

I tried to configure a single 3550 as NTP peer for the three clusters and they work fine with the peer (which get's its time from the very same NTP server).

Debugging NTP didn't make me happy at all as I only saw the clusters prefering the peer before the cluster.

All switches use IOS version 12.1(22)EA4. Searching CCO didn't gave me a clue where to look next. So I try asking you. TIA for any help :-)

Ciao.

Michael Brausen

[Mark Turpin]

Did this config ever work?

Are you using ntp peer's or ntp server x.x.x.x ?

Can you post configs of a working switch and a non-working switch?

Can you post your debugs?

I've had little luck with peers, but server configurations always work.

http://www.cisco.com/en/US/products/sw/iosswrel/ps1831/products_configuration_guide_chapter09186a00800ca66f.html#wp1004877

Thanks,

-Mark

[m.brausen]

Hi,

sorry for my late reply, there were things more urgent :-(

>Did this config ever work?

Yes, I'm sure it did with an earlier IOS Version.

>Are you using ntp peer's or ntp server x.x.x.x ?

I use ntp server and I 'prefer' my primary NTP server.

I updated all switches to IOS Version 12.2(22)EA5 - with no luck.

During debugging I noticed a strange behavior on the three clusters. The cluster uses the cluster's NAT address to send NTP requests instead of the address of the management vlan. And my NTP server resides in a different subnet and for that knows no route back to the cluster. If this is a misbehavior of the IOS or due to my limited understanding of clustering cisco switches, I do not know.

When I configure my routing switch as alternate NTP server all works fine - as it resides in the same subnet. For the moment I can live with that workaround as it adds only one more hierarchy level to NTP which is just fine.

Mark, thank you for your time :-)

Ciao.

Michael