Error Message

LLT INFO V-14-1-10035 timer not called for 105 ticks

The Low Latency Transport (LLT) driver schedules a timer (basetimer) through the kernel function timeout(9F).   The basetimer is 50ms by default on VCS 5.0 and above.   The basetimer was 100ms by default in older version of VCS, e.g. 4.1 and below.   The setting of basetimer can be checked with lltstat -t command, for example,

# lltstat -t
LLT kernel tunables:
    maxnids: 32
    maxports: 32
    nominpad: 0
    basetimer: 50000       <<< 50000 microseconds = 50 milliseconds (ms)

If the kernel doesn't call the LLT timer handler llt_timer_handler() for more than 1 seconds (the kernel is supposed to call the llt_timer_handler() in 50ms), LLT driver will log a system message.   For example,

        LLT INFO V-14-1-10035 timer not called for 105 ticks

Each system clock tick is 10ms. The above message means that the llt_timer_handler() was not called after 1.05 seconds.

One possible reason that the kernel doesn't call the llt_timer_handler() in 50ms as specified is because of Oracle CR6958068 "Nehalem deeper C-states cause erratic scheduling behaviour".

Oracle advised that the following kernel parameter should be set to fix the issue.

/etc/sytsem:

set idle_cpu_no_deep_c=1

Applies To

Solaris 10 x86

Veritas Cluster Server (VCS)