TIMED(8) | System Manager's Manual | TIMED(8) |
timed | [-dMt] [-F host ...] [-G netgroup] [-i network | -n network] |
The following options are available:
The -n and -i flags are mutually exclusive and require as arguments real networks to which the host is connected (see networks(5)). If neither flag is specified, timed will listen on all connected networks.
A timed running without the -M nor -F flags will always remain a slave. If the -F flag is not used, timed will treat all machines as trustworthy.
The timed utility is based on a master-slave scheme. When timed is started on a machine, it asks the master for the network time and sets the host's clock to that time. After that, it accepts synchronization messages periodically sent by the master and calls adjtime(2) to perform the needed corrections on the host's clock.
It also communicates with date(1) in order to set the date globally, and with timedc(8), a timed control utility. If the machine running the master becomes unreachable, the slaves will elect a new master from among those slaves which are running with at least one of the -M and -F flags.
At startup timed normally checks for a master time server on each network to which it is connected, except as modified by the -n and -i options described above. It will request synchronization service from the first master server located. If permitted by the -M or -F flags, it will provide synchronization service on any attached networks on which no trusted master server was detected. Such a server propagates the time computed by the top-level master. The timed utility will periodically check for the presence of a master on those networks for which it is operating as a slave. If it finds that there are no trusted masters on a network, it will begin the election process on that network.
One way to synchronize a group of machines is to use ntpd(8) to synchronize the clock of one machine to a distant standard or a radio receiver and -F hostname to tell its timed to trust only itself.
Messages printed by the kernel on the system console occur with interrupts disabled. This means that the clock stops while they are printing. A machine with many disk or network hardware problems and consequent messages cannot keep good time by itself. Each message typically causes the clock to lose a dozen milliseconds. A time daemon can correct the result.
Messages in the system log about machines that failed to respond usually indicate machines that crashed or were turned off. Complaints about machines that failed to respond to initial time settings are often associated with “multi-homed” machines that looked for time masters on more than one network and eventually chose to become a slave on the other network.
The protocol is based on UDP/IP broadcasts. All machines within the range of a broadcast that are using the TSP protocol must cooperate. There cannot be more than a single administrative domain using the -F flag among all machines reached by a broadcast packet. Failure to follow this rule is usually indicated by complaints concerning “untrusted” machines in the system log.
R. Gusella and S. Zatti, TSP: The Time Synchronization Protocol for UNIX 4.3BSD.
May 11, 1993 | NetBSD 6.1 |