Sometimes the node crashes before VCS is able to write the Diskgroup agent message to engine.log. You can identify if this was caused by PanicSystemOnDGLoss with the following steps.

1. Identify stack back trace for the panic, if it was a panic by the Diskgroup agent we will see

bash-2.05$ mdb -k unix.0 vmcore.0
> $c
vpanic(11750c0, ff3f2e50, 3, ff3f37e0, a, ffbff69c)
kadmin+0x560(5, 0, 0, 30077f0d570, 5, 0)
uadmin+0x104(5, 0, 0, 10034, 0, ff3dec0c)
syscall_trap32+0xa8(5, 0, 0, 10034, 0, ff3dec0c)

2. The Diskgroup agent runs uadmin, this process can be identified as follows:

> ::ps -f !grep uadmin
R 12446 10816 10816 10748 0 0x00004008 00000300e0a2e018 /usr/sbin/uadmin 5 0

3. Identify the parent id which will be the Diskgroup Agent. 

> ::ps -f !grep 10816
R 10816 10748 10816 10748 0 0x02004008 00000307b4e261a0 /opt/VRTSperl/bin/perl -I /opt/VRTSvcs/bin/DiskGroup/../../lib -S /opt/VRTSvcs/
R 12446 10816 10816 10748 0 0x00004008 00000300e0a2e018 /usr/sbin/uadmin 5 0
Z 12418 10816 10816 10748 0 0x00004008 000003005a308b48 /usr/sbin/vxdg -q list

 

 

How to identify a PanicSystemOnDGLoss in a Solaris crash dump