Hello,

I just noticed that we have a few servers that are badly drifting out of time. I'd like to have them sync up to public time server to get their time. My questions are:

1. How do you setup a server to query another NTP server. Do I edit the /etc/ntp.conf file?

2. Not sure how this works, but can you set up a priority to have servers fall back to another time server if the original one isn't reachable?

3. Currently I have the below configured in my ntp.conf, can someone please help me understand this?

server 127.127.1.0 # local clock
fudge 127.127.1.0 stratum 10

4. Also what is a driftfile?

Yes. And then, depending on how far they're out of whack,
either run an "ntpdate server" manually, and then start the
service, or just start the service. Also make sure it (ntpd)
will be started on each boot (how depends on your distro).

Yes. Same file. Just give the machines different "strata".


Cheers,
Tink

1. Yes. Good servers to use are listed at http://www.pool.ntp.org/en/. The names listed are, where possible (usually the case), virtual names that resolve to a "round robin" of servers. Depending on which country you are in you probably need only the name for your country. If the country does not have many NTP servers, list the continent's name, too. Thus, being In India, I configure /etc/ntp.conf like this
   Code:

   # Time sources
   server in.pool.ntp.org
   server asia.pool.ntp.org
   server 127.127.1.0
   fudge 127.127.1.0 stratum 10

2. This is addressed by the pool.ntp.org name being a round robin and listing the continent's name after the country's name in case none of the country's NTP servers are available.
3. AFAIK, the first line allows ntpd to use the hardware clock; the second makes the hardware clock a last resort (most public NTP servers are stratum 2 or 3, meaning they get their time directly from a first quality server or from a stratum 2 server. Thus "stratum 10" is a way down the preference order.
4. AFAIK, a driftfile is used to record how the local clock has drifted from time as served by NTP. NTP is internally very complex and subtle with a lot of thought and experience in its design to provide a best possible service; knowing how the local clock behaves is taken into account, for example to correct the clock if no time servers can be contacted or to determine how often time servers need to be consulted.

ntpd is designed to minimise resource usage so does not often check time servers. After configuring it you need to wait (say 10 to 30 minutes) before checking the logs to see if it is working. This can mean that the relevant messages are widely spaced in the logs. Here's a command that works on Slackware 13.0 to search the logs. It is to be run with /var/log as the current directory; it may need to be modified for other distros