m (→date still stays seconds/minutes wrong: it's offset :))
Latest revision as of 11:40, 26 April 2011
Dovecot and OpenLDAP may react very ugly when your computer-time is not correct.
OpenLDAP needs correct time-values if you want to implement replication. And dovecot simply stops with an Error "time moved backwards"
This problem may be seen especially on virtualized Hosts (like ESX/ESXi/Xen/KVM/...). But also "real server" may benefit of correct time-settings
 current time
if you want to know the current time on your server just type in
compare the result with your own watch :)
The NTP-daemon is not installed by default on every system.
on debian you can install it using
# apt-get install ntp
on CentOS install the package using
# yum install ntp
afterwards you need to check the config-file for multiple lines starting with
... server HOSTNAME-OF-NTP-SERVER ... ...
extend these lines to read like that
... server HOSTNAME-OF-NTP-SERVER .... minpoll 4 maxpoll 10 ...
the "maxpoll" interval is quite important: it tells ntp how often to talk to the ntp-server
the last number 10 (= every 1024 sec) is quite reasonable on real hardware
maybe you want to set it to 8 in virtual environments
please do not set it below 6 (= every 64 sec) as this might be seen as abusive
if you really want to have a very good time sync just use more server as reference
 start/restart the service
afterwards remember to restart your ntp-daemon debian
# /etc/init.d/ntp restart
# /etc/init.d/ntpd restart
 date still stays seconds/minutes wrong
don't worry - this is a feature to keep applications like dovecot alive. The time will be synced in the background and "slowly" adjusted.
# ntpq -n -p remote refid st t when poll reach delay offset jitter ============================================================================== +XX.XX.XX.XX .ATOM. 1 u 2 64 377 27.695 -0.217 0.518 *XXX.XXX.XXX.XXX .ATOM. 1 u 44 64 377 27.985 0.117 0.381 +X.X.X.X .ATOM. 1 u 32 64 377 27.346 0.046 0.273
the row "offset" gives you the difference that needs to be catched up. Just leave it and recheck after a cup of coffee :)
 add to default runlevels
remember to add ntp to default runlevels
on debian this is done automaticaly
for CenOS type
# chkconfig --level 345 ntpd on