After running quick test in sandbox (measure offset of node 2,3 in node 1), I do not see great differences between controller instances before test starts (max=37 msec in 3 runs):
Run 1: 03:15:55 + ssh 10.29.15.1 '/usr/sbin/ntpdate -q 10.29.13.206' 03:16:01 server 10.29.13.206, stratum 2, offset 0.036700, delay 0.02663 03:16:01 19 Apr 03:16:01 ntpdate[11394]: adjust time server 10.29.13.206 offset 0.036700 sec 03:16:01 + ssh 10.29.15.1 '/usr/sbin/ntpdate -q 10.29.13.240' 03:16:08 server 10.29.13.240, stratum 3, offset -0.002716, delay 0.02681 03:16:08 19 Apr 03:16:08 ntpdate[11408]: adjust time server 10.29.13.240 offset -0.002716 sec Run 2: 03:30:38 + ssh 10.29.14.79 '/usr/sbin/ntpdate -q 10.29.12.178' 03:30:44 server 10.29.12.178, stratum 3, offset -0.003602, delay 0.02641 03:30:44 19 Apr 03:30:44 ntpdate[11377]: adjust time server 10.29.12.178 offset -0.003602 sec 03:30:44 + ssh 10.29.14.79 '/usr/sbin/ntpdate -q 10.29.12.173' 03:30:50 server 10.29.12.173, stratum 2, offset -0.003667, delay 0.02621 03:30:50 19 Apr 03:30:50 ntpdate[11391]: adjust time server 10.29.12.173 offset -0.003667 sec Run 3 03:43:50 + ssh 10.29.14.15 '/usr/sbin/ntpdate -q 10.29.14.150' 03:43:56 server 10.29.14.150, stratum 2, offset 0.012081, delay 0.02626 03:43:56 19 Apr 03:43:56 ntpdate[11396]: adjust time server 10.29.14.150 offset 0.012081 sec 03:43:56 + ssh 10.29.14.15 '/usr/sbin/ntpdate -q 10.29.14.110' 03:44:02 server 10.29.14.110, stratum 2, offset 0.020222, delay 0.02615 03:44:02 19 Apr 03:44:02 ntpdate[11411]: adjust time server 10.29.14.110 offset 0.020222 sec I can also push some patch to monitor this for a while. BR/luis > On Apr 18, 2017, at 5:52 PM, Luis Gomez <ece...@gmail.com> wrote: > > Thanks Andy, I will probably do some simple test in sandbox to check clock > synchronization. I am also reading [1] configuring NTP in both host and > guests can help with synchronization, do you know if hosts in rackspace use > hwclock for system time and NTP is enabled on them? > > BR/Luis > > [1] > http://stackoverflow.com/questions/26005056/synchronizing-clock-time-in-host-and-guest-under-kvm > > <http://stackoverflow.com/questions/26005056/synchronizing-clock-time-in-host-and-guest-under-kvm> >> On Apr 18, 2017, at 10:09 AM, Andrew Grimberg via RT >> <helpd...@opendaylight.org <mailto:helpd...@opendaylight.org>> wrote: >> >> On 04/18/2017 09:52 AM, Luis Gomez via RT wrote: >>> Forwarding the sync question helpdesk as I am not sure how the VMs get the >>> time. >> >> Instances get their time initialized by the underlying host upon boot. >> In many cases their "hardware clock" time may be hard stepped each time >> the instance gets sliced in to get an operation turn on the host. >> >> It's possible to configure NTP on instances, but it's usually considered >> non-helpful in general as the time is still going to skew and the amount >> of skew really depends upon the load of the host system. >> >> -Andy- >> >>>> Begin forwarded message: >>>> >>>> From: Robert Varga <n...@hq.sk <mailto:n...@hq.sk>> >>>> Subject: [integration-dev] Time sync between CSIT nodes >>>> Date: April 18, 2017 at 4:41:31 AM PDT >>>> To: "integration-...@lists.opendaylight.org >>>> <mailto:integration-...@lists.opendaylight.org>" >>>> <integration-...@lists.opendaylight.org >>>> <mailto:integration-...@lists.opendaylight.org>> >>>> >>>> Hello, >>>> >>>> do the nodes participating in 3node cluster tests have their clocks >>>> synchronized? >>>> >>>> The logs in https://bugs.opendaylight.org/show_bug.cgi?id=8075 >>>> <https://bugs.opendaylight.org/show_bug.cgi?id=8075> seem to >>>> have some slight skew, but it's hard to tell whether it is natural lag >>>> in local event detection or the nodes have different times. >>>> >>>> If not, would it be possible to establish NTP peering between the nodes, >>>> make sure the clocks are synchronized to a reasonable extent and output >>>> 'ntpq peers' at the end of the test? >>>> >>>> Thanks, >>>> Robert >>>> >>>> >>>> _______________________________________________ >>>> integration-dev mailing list >>>> integration-...@lists.opendaylight.org >>>> <mailto:integration-...@lists.opendaylight.org> >>>> https://lists.opendaylight.org/mailman/listinfo/integration-dev >>> >>> >> >> -- >> Andrew J Grimberg >> Lead, IT Release Engineering >> The Linux Foundation >> >> >> <Mail Attachment> >
_______________________________________________ controller-dev mailing list controller-dev@lists.opendaylight.org https://lists.opendaylight.org/mailman/listinfo/controller-dev
