Opened 10 years ago

Closed 9 years ago

#11 closed defect (fixed)

bug with peer time sync

Reported by: anonymous Owned by:
Priority: minor Milestone: 0.7.11
Component: router/clock Version:
Keywords: Cc:
Parent Tickets:

Description

i recently discovered a bug with the time syncronization with peers.
every 2h or so i got an error message like this and all peers where dropped:
ERROR [Timestamper ] net.i2p.util.Clock : Ignoring clock offset 111458ms (current 197940ms) since it would increase peer clock skew from 1s to -85s. Broken server in pool.ntp.org?

my clock is indeed not correct, but i think thats what the ntp servers are for.
after setting router.clockOffsetSanityCheck=false the error didn`t occure again. at least until now after more than a week uptime (thanks complication ;))

this bug is reproducible (withturning clockOffsetSantiyCheck back on).

For discussion see: ​http://zzz.i2p/topics/149

Subtickets

Change History (4)

comment:1 Changed 10 years ago by welterde

  • Milestone set to 0.6.5

comment:2 Changed 10 years ago by welterde

  • Component changed from router/general to router/clock

comment:3 Changed 10 years ago by welterde

  • Priority changed from major to minor

comment:4 Changed 9 years ago by zzz

  • Resolution set to fixed
  • Status changed from new to closed

Maybe fixed, maybe not.

The clock will only make major changes shortly after startup. But by reducing the minimum change from 10s to 5s, I hope that the code will now keep up with a clock that is way-off. I think what was happening was the clock would get too far off before it checked again.

Perhaps a "real" fix is to implement "real" NTP, that slews the clock gradually, instead of simply making (or not making) step changes.

However, if your system clock glitches every two hours, I don't know how to fix that. Run NTPD.

Note: See TracTickets for help on using tickets.