TOD clock synchronization

6.10 LPAR mode

Your Linux® instance might be part of an extended remote copy (XRC) setup that requires synchronization of the Linux time-of-day (TOD) clock with a Coordinated Timing Network (CTN).

Linux in LPAR mode supports server time protocol (STP) based TOD synchronization. For information about STP, see www.ibm.com/systems/z/advantages/pso/stp.html. Use the lsstp command to display the STP configuration for your Linux instance.

Attention: To avoid hanging I/O operations on XRC-enabled DASD, be sure that a reliable timing signal is available before enabling clock synchronization.

STP support is included in the Linux kernel. No special build options are required.

Note: STP synchronizes leap seconds with a better resolution than Network Time Protocol (NTP). With STP enabled, do not use NTP daemons like chrony or ntpd.

How STP synchronization works

With STP enabled at boot time, STP synchronizes the TOD clock of a Linux instance with the STP timing network during the boot process. STP then steers the TOD clock to keep it in sync with the network. This synchronization is driven by STP, without active participation of the Linux kernel. You cannot enable STP for KVM guests, but KVM hosts pass their synchronized TODs on to their guests.

In contrast, the Linux kernel takes an active role if the TOD clock gets out-of-sync with the timing network. An out-of-sync situation usually occurs when STP is enabled on a running Linux instance. To bring the TOD clock back into sync, STP notifies the Linux kernel through a sync check. The TOD clock then leaps to the corrected time. Linux now shields applications from inconsistent time stamps by gradually steering the values returned by gettimeofday() towards the corrected TOD. Such corrections do not feed through to KVM guests, which remain out-of-sync with their host and with the timing network.