A Solaris system panics when it encounters an unrecoverable software or hardware error. When a panic occurs, it saves data about the state of the system and then reboots. A crash dump is created so that an Oracle engineer can determine the reason for the panic. The post describes a manual way to generate a crash dump in case of a LDOM.
From Control domain
A panic signal can be sent to guest ldom (e.g. ldom01) from the control domain.
control-domain # ldm panic/panic-domain ldom01
From Guest ldom condole
Another way to do this is through VNTSD options. telnet to the port on which ldom vntsd service is running to connect to the console of the hung ldom.
# ldm list NAME STATE FLAGS CONS VCPU MEMORY UTIL UPTIME primary active -n-cv SP 8 4G 0.3% 8h 46m ldom01 active -n--- 5000 8 2G 48% 1h 52m
control-domain # telnet 0 5000
Once you are connected to the console of the guest ldom, press
// on LDom console via telnet (vntsd) telnet> ~# - Send break ~^B - Send alternate break ~. - Exit from this console ~w - Force write access ~n - Console next ~p - Console previous ~? - Help
As seen in the help there are 2 options to send the break signal and generate core dump :
~# ( Tilde / hash Sign ) Send break. ~^B ( Tilde / CTRL-B ) Send alternate break (If "KEYBOARD_ABORT=alternate" has been set in the /etc/default/kbd file within the guest domain)
Select the sync option to generate the crash dump :
Debugging requested; hardware watchdog suspended. c)ontinue, s)ync, r)eset? s