I came across a weird behavior on one of my clients sites. Thought I may ask
here to see if someone knows or seen this behavior.
All of the Linux servers are using MS DC NTP (Win 2003 Server Ent. 64). I know
this is not the best design, but for something as simple as this - MSDC NTP
seemed to be sufficient in my clients opinion.
The layout/design:
Site1 NTP Servers
ntp01-site1
ntp02-site1 (master host that gets its time from pool.ntp.org, everyone else
gets its time from this host)
Site2 NTP Servers
ntp01-site2
ntp02-site2
The problem:
Now I'll be speaking Microsoft (pardon my poor language as I was told this part
by windows engineer). A month go, FISMO role for NTP was change from
ntp02-site1 to ntp01-site1 . ntp01-site1 could not reach pool.ntp.org and now
functions as local time server (yes I know its bad).
While ntpq -q show that host was associated to NTP servers and time drift is
noticed in the offset, ntp did not sync the time and let the time drift.
[root@host01-site2 ~]# ntpq -p
remote refid st t when poll reach delay offset jitter
==============================================================================
ntp01-site2 ntp01-site1 2 u 13 64 377 0.184 23833.3 11.885
ntp02-site2 ntp01-site1 2 u 46 64 377 0.199 23824.1 12.399
Somehow, this broke the NTP sync in my environment for linux hosts but not
windows. No errors anywhere, ntpd runs fine - communicates to ntp serversm I
can see the drift but ntpd never resets the time.
The sample of my ntp.conf file for site2 look like this:
tinker panic 0
server ntp01-site2
server ntp02-site2
restrict default kod nomodify notrap nopeer noquery
restrict 127.0.0.1
driftfile /var/lib/ntp/drift
broadcastdelay 0.008
It would make sense if Site1 hosts stopped functioning because local NTP server
was unable to communicate upstream to ntp.org, but why would site2 be affected.
After all it gets its time from site1.
What is the problem? Why would ntp not function in isolated mode like this? Am
I missing an option in ntp.conf that would avoid this behavior? Is this a
general design flow or MS bug?
Any help is appreciated,
Thanks
ilya
_______________________________________________
rhelv5-list mailing list
rhelv5-list@redhat.com
https://www.redhat.com/mailman/listinfo/rhelv5-list
