NTP with no outside source avalilable

figadiablo · 04-07-2005, 12:00 AM

Ok, hi.

I have a small network. In a lab, with no connection whatsoever to the outside. Because of the experiments we are running, I need to have all the computers' clock syncronized. All application log files' date and time should match on all computers.

According to the help files from the ntp site, and from reading the ntp.conf file, using this, aka. "Undisciplined Local Clock" settup:

Code:

server 127.127.1.0
fudge 127.127.1.0 stratum 3

my computer should syncronize with itself. But it is not happening. So far I haven't found a real guide to set up a server, this way. But there should be a way.

If I edit the ntp.conf file, and then start it up, via ntpd, I get no error messages or nothing, but the ntpdate command fails to syncronize. Is the ntpdate command supposed to work with itself.

I did notice this:

Code:

bash-3.00# ps aux | grep ntp
root      6756  0.0  0.3  3588 3584 ?        SLs  01:26   0:00 ntpd
root      6816  0.0  0.0  1688  600 pts/0    R+   01:49   0:00 grep ntp
bash-3.00# ntpd
ntpd     ntpdate  ntpdc    
bash-3.00# ntpdc
ntpdc> sysinfo
system peer:          LOCAL(1)
system peer mode:     client
leap indicator:       00
stratum:              4
precision:            -18
root distance:        0.00000 s
root dispersion:      0.01250 s
reference ID:         [127.127.1.1]
reference time:       c5ff4838.00fb00bc  Thu, Apr  7 2005  1:48:40.003
system flags:         monitor ntp kernel stats 
jitter:               0.000000 s
stability:            0.000 ppm
broadcastdelay:       0.003998 s
authdelay:            0.000000 s
ntpdc> q
bash-3.00# ntpdate 
 7 Apr 01:49:29 ntpdate[6818]: no servers can be used, exiting
bash-3.00#

notice the stratum line:

Code:

stratum:              4

Could it be that it is working and the other computer should sync with this one. I don't have another computers withme right now, But I do at work. I though that maybe someone could shine some light in here.

Thanks,
Figa

TigerOC · 04-07-2005, 02:33 AM

I would say that you need to use the localhost address for server i.e. 127.0.0.1 Technically 127.127.1.0 does not exist. With regard to the different stratum layer it will always indicate 1 level below the referenced server. Using localhost as the time source should work. You only then need to configure ntp clients on the other boxes to user server <ip_address of the primary server>. Dont not forget to setup the drift and log files so that ntp can auto-regulate itself and the same goes for the other boxes.

TigerOC · 04-07-2005, 03:45 AM

Having given your situation some thought here is what I would do. You need a median time so lets say you have 4 boxes, A,B,C and D. Start with A. Configure ntp.conf like so;

server <ip-address of B>
server <ip-address of C>
server <ip-address of D>

moving to B;

server <ip-address of A>
server <ip-address of C>
server <ip-address of D>

You get the drift. Providing you have the drift files setup on each of the boxes they will probably all sync to the box with the least drift and therefore the one with the most stable time oscilator. It may take a few hours for the time to stabilise. You can see how each is getting on with the cli query ntpq -p and you will soon see which box they have focussed on.