[rescue] question about crashdump device in mirrored swap configuration

Sandwich Maker adh at an.bradford.ma.us
Fri Jul 10 11:51:39 CDT 2009

" From: "Curtis H. Wilbar Jr." <rescue at hawkmountain.net>
" In a mirrored swap configuration (under Solaris 8), should the dump 
" device in /etc/dumpadm.conf be changed to the metadevice for the swap
" space.... it is currently set to /dev/dsk/c0d0s1.  Problem is the
" machine is rebooting randomly....

i -think- it should be set to the raw device, as i'm not sure it's done
after the os/kernel is up.

you could try dumpadm -d swap and see what it does.  the manpg also
says dump can go to a preallocated file in the crash fs if i
understand correctly.  does the system have space?

" if it is panicing it is not leaving a crash dump in the 
" /var/crash/<hostname> directory.

you have to enable savecore to get this.

" Server is remote, and I believe it is a hardware issue (it is a 'cheap' 
" x86 box (yes, it is Solaris 8 for x86 (a client's box)).  It ran stable
" for ages... so I'm guessing more than likely hardware (fans, ps, memory,
" etc), but I was unable to find an answer about if dumpadm.conf should
" be updated.
" If not... then what does the metadevice do when the system boots and the 
" swap space mirror components are no longer in sync (due to a crashdump
" being written to one 'half' of the mirror) ?

for a fs like root, if it's in the side you booted from it'll be synced
to the other.  since the os can run without any swap, it might ignore
junk not written to the md after the swapon.  otoh if you -tell- dump
to use the md...

btw, i've found that if you set maxphys -and- maxcontig way up, syncs
run faster.
Andrew Hay                                  the genius nature
internet rambler                            is to see what all have seen
adh at an.bradford.ma.us                       and think what none thought

More information about the rescue mailing list