[Orca-dev] orcallator needing to be root (was Re[2]: orca-ing data already in a database? (fwd))

[Rusty Carruth]

I sent this to Blair privately - and he said I might just as well
send it to the group, so here it is!

Since I'm now sorta developing on orca, I suppose I should switch this 
discussion to the developers list - I just joined there also and am
sending this there as well.  Unless informed otherwise, I'll move all
my discussions about this over to -devel*.

Anyway:

Basically, I want to make orcallator work without having to be root.

Blair Zajac <blair at orcaware.com> wrote:
> Rusty Carruth wrote:
> ...
> > > would be too hard to figure out.  You'll probably add #ifdef 0 #endif around
> > > pieces of code until it doesn't complain.
> > 
> > Hmm.  Why didn't *I* think of that?  (excuse - because this performance
> > logging
> > is a side project ;-)
> > 
> > I assume you'd like the patches when I'm done?
> 
> Of course :)

Now, THIS is strange.  The only reason it won't run as root is because,
on my Sun 2.5.1 machine, the number of disk devices is wrong.

Here's the output without my patch:

Writing data into /export/home/orca/var/orca/orcallator/msfree/
Starting logging
Warning: Cannot init kvm: Permission denied
Fatal: subscript: 2 out of range for: GLOBAL_disk_info

Now, with my patch:

Writing data into /export/home/orca/var/orca/orcallator/msfree/
Starting logging
Warning: Cannot init kvm: Permission denied
adding raw disk.  short name: dad0
adding raw disk.  short name: atapicd2
adding raw disk.  short name: fd0
raw disk count is 3, changed to 2
I is now 0
I is now 1
adding raw disk.  short name: fd0
raw disk count is 3, changed to 2
I is now 0
I is now 1

Other than the junky output, I've simply subtracted one from
the number of raw disk devices in raw_disk_map!  And it seems to be working.

I'm going to go over to machines in the other building (about 700 miles away!)
and try them out to see if I can figure out if there are any other 
bad devices...

> > (And of course it will be fixed so that it uses effective uid to
> > choose to do those things or not, and not a #ifdef 0! ;-)
> 
> I would think you would want to patch orcallator/start_orcallator.sh.in
> and either define or undef a CPP define that would be set to either
> UID_IS_ROOT=
> 
> or
> 
> UID_IS_ROOT=-DUID_IS_ROOT
> 
> and then pass to se as:
> 
> se $UID_IS_ROOT orcallator.se

hmm.  You'd rather not autodetect this?  (ASSUMING that its not going to be
something that has to be hand-tweaked!)

rc


<<  End forwarded message