On Mon, Jul 25, 2011 at 8:51 AM, Paul Loy <ketera...@gmail.com> wrote:
> As I understand it, this will not quarantee that they are millisecond > accurate which is what you need for Cassandra to use the correct commit. > We've seen problems in production and had to rearchitect parts of the system > due to this even though all servers are NTP synched. > > http://www.endruntechnologies.com/faq.htm#How_accurate_is > > > Perhaps you'd do well to do some more intensive NTP configuration. Something like: outside server(s) <--- 2 local NTP machines (time[01].mydomain) <---- X number of cassandra servers over lan assuming your lan is speedy enough, you should be able to get to the .5 MS range of sameness (which might still be 100MS off 'real' time, but I don't think you'd really care about that). it might even be worth creating your own NTP LAN that connects all the machines on an interface separate from Data & Management. -- Even the Magic 8 ball has an opinion on email clients: Outlook not so good.
