DUAS: crash uxioserv after Solaris Intel 11.4 update

Products

• CA Automic Dollar Universe

Issue/Introduction

After having applied a Solaris 11 Intel patch 11.4.19.3.0 or 11.4.26.0.1.75.4, the uxioserv process of the Dollar Universe 5.3 or 6.x nodes running on the system crashes with a signal 6 generating a core file.

|FATAL|X|IO |pid=pid.tid| k_process_signal          | *** execution handler: SIGNAL=(6) PID=(pid) PPID=(1) GPID=(pid)

The errors that appear in universe.log at the time of the crash is:

ASSERTION CAUGHT: ts->tv_nsec >= 0 (0xffffffff8000007c >= 0x0), file: /builds/s11u4/11.4.19.0.1.3.0.server/usr/src/common/hrt/i386/timestamp_cmn.c, line: 167

A more recent  Solaris 11.4 patch provided by Oracle produces a similar error on a different line:

ASSERTION CAUGHT: ts->tv_nsec >= 0 (0xffffffffaf28c47f >= 0x0), file: /builds/s11u4/11.4.26.0.1.75.4.server/usr/src/common/hrt/i386/timestamp_cmn.c, line: 202

 

The core analysis shows that the process crashed when calling a system C library called libc_hwcap1.so.1:

$c
libc_hwcap1.so.1`__lwp_park+0x19(80d3174, 80f5248)

Which points to /lib/libc.so.1

/usr/lib/libc/libc_hwcap1.so.1
                        33G   4.7G        28G    15%    /lib/libc.so.1

 

Environment

Release : affects both 5.x and 6.x

Component : DOLLAR UNIVERSE

Cause

Issue seems to be caused by a bug on the System library /lib/libc.so.1

 

Resolution

Workaround: the node automatically restarts then the IO process crashes

Solution:

We suspect there must be a bug on the version of Solaris 11 update 4  library /lib/libc.so.1, please open a case with Oracle Support to get a fix for it.