NTP (Network Time Protocol)

Basics

------

- user space daemon is "ntpd" running on udp port 123

- ntpd updates the system clock

- system clock is a software clock running in the kernel w/c is more accurate

  compared to RTC

- RTC is the hardware clock whose value varies depending on the temperature

  (real time clock)

What is Drift File?

-------------------

- stores the frequency offset between the system clock running at its nominal

  frequency and the frequency required to remain in synchronization with UTC

- used to adjust system clock every time ntpd start

- located in: /var/lib/ntp/drift (directory must be writable by NTPD)

- entry on config file: driftfile /var/lib/ntp/drift

Time on Virtual Machines

------------------------

- they have no hardware clock w/c is a more accurate time source

- load of hosts (ESX) where they resides affects the timekeeping (e.g higher

  load slows down clock ticks)

NTP Strata

----------

stratum 0	atomic clocks, GPS, mobile phone systems
stratum 1	computer with radio clock, GPSclock, or atomic clock attached
stratum 2	Reads from stratum 1; Serves to lower strata
stratum 3	Reads from stratum 2; Serves to lower strata
stratum n+1	Reads from stratum n; Serves to lower strata
stratum 15	Reads from stratum 14; This is the lowest stratum
stratum 16	unsynchronized state

How ntpd chooses a time source?

-------------------------------

1. it will check which is the closest (e.g least delay, offset, jitter)

2. it will check the stratum

3. it will check "prefer" directive in /etc/ntp.conf

NTP configuration file: /etc/ntp.conf

-------------------------------------

assigning a time source	## basic server [options] ## assigning from an ntp pool server 0.rhel.pool.ntp.org iburst
restricting access	## don't allow IPs on these list to modify your config file and query ## your NTP server restrict 192.168.1.0 mask 255.255.255.255 nomodify noquery ## makes sure localhost has full permissions w/o any restricting keyword restrict 127.0.0.1
misc	## fails back to local clock if no time source is available server 127.127.1.0 fudge 127.127.1.0 stratum 10

Some facts about leap seconds

-----------------------------

- affected systems are only those who use NTP

- by default, systems not running NTP are not affected

- leap second might make the system (or the application running inside) crash

- when a leap second is inserted, kernel prints it to /var/log/messages

- Cron jobs scheduled on 00:00 UTC won't be launched twice during leap second

Action Plans to do for Leap Second (from Redhat support)

--------------------------------------------------------

1) You can stop ntpd before June 30th, and then start it again when it is more

   convenient. Once it is started, it will sync with the NTP servers.

2) Use ntpd in slew mode (-x option). With this configuration the time won't be

   stepped, instead it will be gradually adjusted.

Commands

--------

Displaying

## checks time servers (detailed status) ntpq -p COLUMNS: remote and refid = remote NT P server, and its NT P server st = stratum of server t = type of server (local, unicast, multicast, or broadcast) poll = how frequently to query server (in seconds) when = how long since last poll (in seconds) reach = octal bitmask of success or failure of last 8 queries         (left-shifted); 377 = 11111111 = all recent queries were         successful; 257 = 10101111 = 4 most recent were successful,              5 and 7 failed delay = network round trip time (in milliseconds) offset = difference between local clock and remote clock         (in milliseconds) jitter = difference of successive time values from server (high               jitter could be due to an unstable clock NOTE: - unreliable time source are those with high delays, offset, and jitters - its current time source is the one with * ## checks for brief ntp status ntpstat ## syncs with ntp server manually ntpdate –u ## queries only; doesn't set the clock ntpdate -q