Hide Forgot
Description of problem: In NTPv3 symmetric active mode the peer is assumed dead (shown as LastRx 10y in chronyc sources output. Version-Release number of selected component (if applicable): chrony-1.25-1.fc14.x86_64 How reproducible: Always; Steps to Reproduce: 1. configure a non-chrony peer in chrony.conf: peer 10.0.0.1 2. do the samew on the peer in ntp.conf: peer 10.0.0.2 3.reload both service instances 4 check chronyc sources output. Will see a line like this: =? 10.0.0.1 0 6 10y 0ns[0ns] +/- 0ns Actual results: the peer assumed dead Expected results: ntp synchronisation Additional info: In a tcpdump you can see the NTPv3 symc_act packets as usual. (Usual for NTPV3 sym_act mode, but unusual otherwise as NTPV3 peers in this mode do not respond to NTP messages directly but each peer on its own schedule. In this mode the proper check against lost packet is if no packet arrived in a full poll interval, not the usual hlf-to two seconds.)
Can you try adding "version 3" to the peer line in ntp.conf? Chrony before 1.26 required the version in NTP packet for peer associations to be 3, in 1.26 this was relaxed and any of 2, 3 or 4 should work.
the ntp4 server was falling back to nptv3 seeing chronyd's ntpv3 packets. As I said the issue is with implementation of NTPv3 symmetric mode. Adding a version 3 option will not change anything, I suppose. Today late evening I can make a try with it of course, as I will have access to the units under test. Do you need detailed explanation of NTPv3 symmetric mode? (The relevant RFC is enough copious and over-sophisticated)
You are absolutely right, (and I was searching in the cause in wrong direction) the version 3 command option in ntp.conf solved the issue. BTW you know any fedora14 backport of the 1.26 chrony?
Ok, thanks. I don't know if someone did a backport to F14. The F15 chrony package uses systemd units, so just recompiling the F15 package on F14 probably won't work. Also, the buildrequirement on pps-tools-devel should be removed.