[rescue] Potential SVM issue in Solaris 10; anybody got a spud?

[Bryan Gurney]

On Mon, 05 Jun 2006 21:41:46 -0400, Skeezics Boondoggle <skeezics at q7.com>  
wrote:

>> On 6/4/06, Bryan Gurney <arb_npx42 at comcast.net> wrote:
>> > What are the best practice rules on logging with UFS, for root
>> > and other partitions?  I booted up the U60 on the 18 GB disk
>> > install of the initial Solaris 10 release, and it only has logging
>> > on the slice I used for /export/home.  One of my other Solaris-
>> > savvy friends says that ideally, logging should be enabled on
>> > all partitions, because otherwise the filesystems aren't
>> > journaled in case of a power outage or spontaneous reboot.
>>
>
> well, in general, yeah... but how much are you _writing_ to your / or
> /etc?  turning off journaling for partitions that are largely readonly
> shouldn't be much of a loss... but then, i'm still used to the pre-bloat
> solaris 7 era, where i could mirror a pair of 9gb drives, create a 1gb /,
> a roomy swap and /var, and still have room left over for a local copy of
> /usr/local if i wanted to (rather than nfs mounting that).  i suppose if
> folks are doing the "modern" thing and creating a swap slice and lumping
> the rest into /, then yeah, you'll probably want to journal it.
>
> sigh.

Well, here's how I partitioned things:

c0t0d0s0: / 				(10240 MB)
c0t0d0s1: [swap] 				(514 MB)
c0t0d0s2: [whole disk]
c0t0d0s3: [partition for metadb] 	(33 MB)
c0t0d0s4: /var				(12800 MB)
c0t0d0s5: [partition for metadb]	(33 MB)
c0t0d0s6: [partition for metadb]	(33 MB)
c0t0d0s7: /export/home			(46351 MB)

Admittedly it's rudimentary, and I'm not 100% happy with it (/usr, /opt,  
and /tmp are in the root partition, wtf?!), but it seems a little sane for  
now.  I also set up three partitions for the metadb's so I can have three  
metadb's if one drive dies.  Unfortunately having three metadb partitions  
limits me to 3 partitions for stuff (since swap and whole disk eat up two  
out of the eight partitions).  This box isn't going to be doing anything  
heavy; my goals are some Apache + PHP + PostgreSQL for developing my  
meager DB skills (which explains /var; 12 GB should be enough to play  
around), and a volume to stash away some backup files from other machines  
(because my experience with DVD readability is less than stellar, and  
because I'm too cheap to buy a good tape drive, and too afraid to buy a  
used DLT8000.  Oh, and I loathe what happens when tape reads go bad; don't  
you?).


> i've been stuck in suse hell at my new job, and the last _significant_
> solaris work i got to do (for money, anyway :-) was solaris 7.

Could be worse; 90% of my job time is spent on NT.  My previous job was  
80% *nix (mixed between Solaris and RHEL), but they wouldn't take me to  
permanent with benies, even after 6 months, and a significant part of my  
time was eaten up with desktop support.

Add onto all this that my original degree was in Computer Engineering, and  
I was initially aiming for some field like microcontrollers, DSP, or  
something like that, but all of my leads fell through, and I was left in a  
post-Titanic-esque dark freezing seawater, searching for ANY job.  I knew  
a thing or two about setting up Linux systems, had tooled around with  
Solaris 9 a bit back in summer 2004, and had plenty of experience setting  
up, administering, and cleaning crapware off of Windoze systems (2000 Pro,  
nothing in the Server realm because you can't go download Win2K server for  
free and tweak around with it).  That became my fallback, and I applied  
for desktop support things, eventually falling on that *nix/windows junior  
admin position.  Then I outgrew the desktop side and wanted to focus on  
servers.  Still not much room for a rookie, but I eventually found  
something.  At least I'm expanding my NT server admin skills portfolio. :-P

I still think now and then of what could have been.  Could I have joined  
in a microcontroller group and gone up from there?  Perhaps a hardware  
systems testing job that could lead more to design positions.  I still  
cringe a little when people say "the software is more important than the  
hardware".  The software would not exist without the host hardware.   
Someone has to make, develop, and enhance that hardware.  Unfortunately,  
the golden days of $60,000 workstations funding research operations; even  
Sun is trimming down a bit now that they're using the Schwartz.  I'm in a  
rather good position now, so I'll probably accumulate some experience, and  
then see if I want to continue in the administration field, or if  
something in hardware/systems design opens up again.  I was in a research  
internship position before that dealt a lot with electronics and  
microcontrollers, and I liked the research environment, as long as there  
was funding.

But anyway...

> uh, anyway, back to the point:  if you still keep a separate /var
> partition (where you should enable journaling) and a small, basically
> read-only / (and you're using cfengine or something similar to manage  
> your
> config files, right? :-) then any sun4u or relatively recent x86 box
> should be able to fsck that in no time flat, and you still can boot from
> either half of the mirror without grief, patch or no patch.

Uh oh, this is probably where my monster / is going to bite me in the  
end.  It's already 33% used up after the initial install.  For my previous  
Solaris 10 installs, I used to separate /opt into its own partition, and  
after seeing how big /usr was, I was going to do the same for that, but  
with my scheme above, I can only have 3 data partitions, total.