Can I ask why the default drift allowed (50 milliseconds) is so small ? . It seems that the default monitor lease is about 5 seconds. And auth ticket expiration time should also be more than 5 seconds. In my practice, even if I’m using NTP, I still got clock skew warnings from time to time. Can I just increase the mon_clock_drift_allowed to 1 second? I assume the impact is when the monitor quorum issue a new voting (one of the monitors is down), there might be two leaders simultaneously working in one second instead of 50 milliseconds ? How to measure the impact if I increase the mon_clock_drift_allowed to one second? van chaofanyu@xxxxxxxxxxx > On Jun 5, 2016, at 10:35 AM, Sage Weil <sage@xxxxxxxxxxxx> wrote: > > On Sat, 4 Jun 2016, Wei Jin wrote: >> In my opinion, if clock drift is too heavy, the lease protocol inside >> paxos won't work. Lease may be timeout and then triggers monitor to >> bootstrap. > > Exactly. > >> You could try to shift clock back on the leader node to see what will happen. >> >> Time check mechanism gives us warning message so that we may take >> actions if possible before things getting worse. > > You can also get into situations where the clients loop trying to get a > new ticket because the one they are issued appears to be expired. > > sage > >> >> On Sat, Jun 4, 2016 at 12:06 PM, Ning Yao <zay11022@xxxxxxxxx> wrote: >>> Hi, all >>> >>> I find that even if I have mon clock skew warning, under most cases, >>> my cluster still can work properly as expected. >>> >>> I go through the code in mon but do not find any related issue without >>> time_check() to update whether clock skew periodically. >>> >>> can any one tell what is the design purpose of the clock skew between >>> mon and when it will really matter the cluster? >>> >>> Thanks a lots! >>> >>> Regards >>> Ning Yao >>> -- >>> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in >>> the body of a message to majordomo@xxxxxxxxxxxxxxx >>> More majordomo info at http://vger.kernel.org/majordomo-info.html >> -- >> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in >> the body of a message to majordomo@xxxxxxxxxxxxxxx >> More majordomo info at http://vger.kernel.org/majordomo-info.html >> >> > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
