>> When run on a node with no RTC, this line causes chrony to disrupt the >> other nodes until the node manages to synchronise with a network node.
> Could you please describe what kind of disruptions you’re seeing? Our Internet gateway synchronised with the NTP pool. There are a number of nodes that are clients of the gateway and they all peer with each other. As expected from Murphy's Law, the node with the least stable clock (massive amounts of jitter) is running Jessie. When the Internet connection goes down, the node with the unstable clock declares itself as a stratum 10 server, and the nodes with stable clocks get jittery time while claiming NTP sync (in the status field of the adjtimex system call). This is fixed by a one-line change to the configuration, granted, but since this default is unexpected, it took me a while to find where the issue came from. (Je ne t'en veux pas trop, mais ne recommence pas ;-) >> I suggest commenting it out in the default configuration. > Well, I’m quite reluctant about introducing such a change to stable as > I suspect some users are relying on this feature. I see that this has been fixed in Stretch, so please feel free to mark this bug as fixed. Thanks for your attention, -- Juliusz
