Re: Are __db.00[123] lock files or not?

[Andre Costa <acosta ar microlink com br>]

On Thu, 15 Jan 2004 16:22:13 -0500
Jeff Johnson <jbj@redhat.com> wrote:

> On Thu, Jan 15, 2004 at 07:21:58PM -0200, Andre Costa wrote:
> > 
> > Please apologize for the confusion.
> 
> No real problem here, it's just that most of a year of my life
> has been watsed with __db* files. That has nothing whatsoever
> to do with you.

I understand, it should be a PITA seeing this surface over and over
again. Hoperfully, this thread will help people to don't ask about it
again since all that needs to be understood is here.

> Try
> 	cd /var/lib/rpm
> 	/usr/lib/rpm/rpmdb_stat -CA
> and watch the output. Run rpm -qa as root, ^Z out, and you
> will se a lock appear. Restart the rpm -qa, run to completion,
> watch the lock disappear.

I had done this another way: I ran synaptic and tried the rpmdb_stat -CA
while it was running. Finally I could see one of these rare creatures ;)

> Repeat, but kill the rpm -qa with kill -9, watch a persistent
> stale lock be left behind.
> 
> Then try running rpm -qa while installing, as that's the "feature"
> that these locks are about, concurrent access.
> 
> In fact, try installing 2 packages simultaneously, that works.
> 
> Can you break concurrent access by installing hundreds of pkgs
> simultaneously? Sure, feel free to do so, if the problem is exactly
> reproducible, I may even have a prayer of attempting a fix.

Now, these seem to be 'serious' fun. I guess I'll pass...

> Anyways, I dinna mean to be gruff, apologies.

No worries, no harm done at all =)

Best,

Andre

-- 
Andre Oliveira da Costa