Time drifting

[swardell]

I cant quite see how to create a trigger that fires when the time on a monitored server (host local time) has drifted by a value x (minutes) from either the Zabbix server time or ideally an NTP Servers time, anyone have any suggestions?

All my servers do a time synch to NTP servers but for some reason I am seeing one drifting and I want to try to catch it

[Alexei]

Have a look at trigger function fuzzytime(), might be handy.

[untergeek]

UserParameter=ntp.offset,/home/zabbix/bin/ntp_test.sh

/home/zabbix/bin/ntp_test.sh

#!/bin/bash

# $Rev:: 691           $:  Revision of last commit
# $Author:: REDACTED     $:  Author of last commit
# $Date:: 2009-12-28 1#$:  Date of last commit

TEST=$(/usr/sbin/ntpq -pn 2>&1 | grep -c "Connection refused")
if [ "$TEST" -gt "0" ]; then
   RETURN=4.99;
else
   RETURN=$(/usr/sbin/ntpq -pn | /usr/bin/awk 'BEGIN { offset=1000 } $1 ~ /\*/ { offset=$9 } END { print offset }')
fi

echo $RETURN

This works for us. We can then define the trigger around the value returned.

[LenR]

{hostname.whatever:system.localtime.fuzzytime(60)} =0

Would trigger if the client time is 60 seconds different from the zabbix server.

We have our own NTP servers, our zabbix server runs an ntp client to keep synced. I have a trigger for 10 seconds of drift between the zabbix server and the ntp servers.

BTW, we had to run NTP clients on vm's and adjust kernel parms for various OS levels to stop time drift.