0

I have two routers that; when first powered up, will initially form full OSPF neighborship with each other and exchange their respective advertised routes with each other. After approximately 2 hours they'll eventually drop their neighborship and routes will be lost. These routers are typically powered off at the end of the day and then brought back up the following day.

I've noticed that their clocks are not in sync. And, to be clear, I'm not referring to whether they are using NTP and/or synchronizing with each other or another NTP server. I simply mean that their time values, when viewing their clocks, do not match. NTP is not being used in this case. Which, I completely understand that; for many reasons, they should be using NTP.

If anyone is curious about what kind of routers these are; they're Linux based (CentOS 6) routers running quagga as their routing daemon.

But with respect to maintaining OSPF neighbor relationships how critical is it that their clocks be close to or in sync with each other? I imagine that there could be some clock drift that occurs when not using NTP. I'm not sure if hold and dead timers could be affected. What other aspects of the OSPF dynamic routing protocol is dependent upon time?

  • What kind of authentication is used? – Zac67 Dec 31 '19 at 18:14
  • 1
    OSPF does not normally depend on time, but not having network device clocks in sync will cause you no end of trouble when troubleshooting and looking at the logs because you really cannot tell what happened when. You should really have all your device times synchronized to a single time zone, typically UTC if you have multiple time zones. That way, all the logs can be checked and will have a common time so that you can tell what happened at what time. – Ron Maupin Dec 31 '19 at 18:18
  • @Zac67 we're using MD5 authentication. – dutsnekcirf Dec 31 '19 at 18:41
  • @RonMaupin, No doubt we need to implement NTP on these devices to help sync logs. We plan to get that in place as soon as we reasonably can. – dutsnekcirf Dec 31 '19 at 18:44
2

OSPF doesn't exchange timestamps, so offset time between routers shouldn't be an issue. MD5 authentication doesn't rely on synchronized time either. Without more details on your config, device time doesn't seem to be your problem.

However, since you might need to rely on synchronized logs (which are a very nice thing, to say the least) or eventually on other forms of authentication (eg. Kerberos), NTP synchronization is very important in general.

Additionally, it's a good idea to design a clean hierarchy for time distribution in your network to avoid internal skew under all circumstances. We're using a single (stratum 1) node to get time over the Internet (from ptbtime.de) and then distribute from there - each node on the HQ net and one (stratum 2) node in all branch networks. All other nodes in each branch (stratum 3) synchronize to their local stratum 2 node. That way, skew in each location and - normally - across locations remains minimal, no matter what kind of errors, link failures etc come up.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.