[ntp:questions] NTP servers redundancy
unruh at wormhole.physics.ubc.ca
Tue Jan 19 20:26:30 UTC 2010
On 2010-01-19, Richard B. Gilbert <rgilbert88 at comcast.net> wrote:
> Rob wrote:
>> Richard B. Gilbert <rgilbert88 at comcast.net> wrote:
>>> The fundamental problem with two servers is this: which one do you
>>> believe when the two differ? You know that at least one of the two must
>>> be wrong but it's impossible to determine WHICH one!
>> This is not relevant when the two servers return the same time within
>> the margin that one considers acceptable.
> It would be wonderful if we could rely on two servers always agreeing!
> The last "NTP Survey" (google for it) found at least one server that
> responded with the wrong year!
>> So when you have two servers that each are synchronized to the same
>> GPS PPS clock, there should be no problem. They return the same time
>> or no response at all (when down) and the clients will lock on either
>> of them without problem.
> And when that "GPS PPS Clock" fails? It should not but that's not the
> same as cannot!
Then you are screwed. Even 10000 servers would not help.
>> When a server returns the wrong time and still says it is synchronized,
>> there is a software bug. This you aren't going to solve with as many
>> servers as you like (all setup the same way). Then you would need to
>> run many servers with different software versions, different OS, different
>> hardware, etc.
>> Probably more trouble than the problem is worth.
More information about the questions