[Pool] Leap second at end of June 2015

Majdi S. Abbas msa at latt.net
Tue Jan 6 17:36:07 UTC 2015


On Tue, Jan 06, 2015 at 12:36:26AM -0800, Ask Bjørn Hansen wrote:
> Yes, sorta. Since last week[1] the system is tracking the information.
> 
> There's an open todo to do the work to track the correct information and 
> email operators when their server is wrong.

	Excellent!

> Until I have more data I don't think it is useful to kick out servers who 
> have it wrong as by the time we know I expect that it's too late to help 
> most(?) clients anyway...

	I wouldn't pull anyone from the pool over it; the worst case,
they go falseticker briefly and recover.  Given the long poll intervals
of a fully synchronized client, only some subset would ever even see
the 1s error, and then the server should step back to reality pretty
quickly.

	In the past I've seen many servers fail to handle the leap
correctly and still recover within 10-15 minutes.  Given the long 
tail of DNS changes, by the time they were actually mostly out of the
pool they'd have recovered.

	--msa


More information about the pool mailing list