Hi,
On Mon, 2014-09-15 at 17:41 +0200, Peter Meerwald wrote:
> Hello Patrik,
>
> connman builds up available timeservers and keeps them in a list, each
> timeservers is tried and used in turn
>
> if a timeserver fails (not resolvable, not in sync, kiss-of-death),
> current timeserver is REMOVED from the list and next one in the list is
> used
>
> is this the intended behaviour?
So far yes. Every 2h ConnMan wakes up and compares the existing
timeserver list against a freshly created one. If the first element of
the current and the fresh timeserver lists differ, the timeserver
resolution is restarted using the fresh list. This way more preferred
timeservers are retried every once in a while should they have been
removed due to connectivity problems. And of course if the connection is
changed, the timeserver list is reset. See commit
3b70a3d5a93fb229c2caacf6b2df43dd9ea5b6bd for details.
> cycling through the timeserver list is a good idea, but why removing them?
> after a temporary network outage, a connection may recover but looses all
> its timeservers
The problem here is to know when to stop and not keep on cycling through
the list banging on timeservers that are not answering. At the same time
more preferred ones should be used over the global ones.
Perhaps a modification could be made so that if the list gets exhausted,
the set of timeservers is retried in something like 5-10 mins?
Cheers,
Patrik
_______________________________________________
connman mailing list
[email protected]
https://lists.connman.net/mailman/listinfo/connman
