[ic] Perl 5.8.0 + Interchange Clustering locking issues

Mike Heins interchange-users@icdevgroup.org
Tue Sep 10 10:52:03 2002


Quoting cfm@maine.com (cfm@maine.com):
> On Mon, Sep 09, 2002 at 11:29:23PM -0400, Mike Heins wrote:
> > Quoting cfm@maine.com (cfm@maine.com):
> > > 
> > > Oh.  Silly me thinking another system looks like ours.  We have
> > > everything but the sessions and tmp in mysql.  Interesting, I've never
> > > looked at tmp except to clean it.  Hmmm, they are all counter files.
> > > I suppose that could be an issue.  OTOH, if they are just counter
> > > files, then move them into sql too.  What are these things?  They are
> > > all just 1 on our system:
> > > #COUNTER-1.0
> > > 1
> > > #COUNTER-1.0
> > > 1
> > > #COUNTER-1.0
> > > Looks like another feature we don't use.
> > > 
> > 
> > No, you use that one if the file even exists. Most of them will be 1
> > because you don't have users coming in with multiple sessions assigned
> > to one IP.
> 
> Thank you Mike.  You didn't even give me the chance this am to look
> at the source  :-)
> 
> It sounds to me that for those using SQL based sessions, putting the
> tmp files in SQL would not be hard.  IF that is in fact the cause of
> the signal error.

We might consider putting that one optionally in SQL anyway -- I think
with 4.9's new ability to define a [counter sql="table:column"], that 
could cause all need for fcntl/flock to go away if SQL sessions are
used...

-- 
Mike Heins
Perusion -- Expert Interchange Consulting    http://www.perusion.com/
phone +1.513.523.7621      <mike@perusion.com>

Be patient. God isn't finished with me yet.  -- unknown