[ic] Verisign, double, tripe charges, orders not going through IC

Ron Phipps interchange-users@interchange.redhat.com
Fri Sep 21 12:58:00 2001


> From: interchange-users-admin@interchange.redhat.com
[mailto:interchange-
> users-admin@interchange.redhat.com] On Behalf Of mheins@redhat.com
> 
> Quoting Ron Phipps (rphipps@reliant-solutions.com):
> > We're back to a system level issue on RedHat Linux.  I have since
> > replaced the v3 Verisign with v2 which was working properly on this
> > server before we moved to IC 4.8.1.  I just now saw the double and
> > triple charge issue which was present in v3.  We are now seeing in
the
> > logs the problem with processing cc transactions as well as the
> > following sporadically:
> >
> > caseetc /c
> > gi-bin/caseetc/process.html Unable to send mail using
/usr/lib/sendmail
> > >
> > > (this contains the email that was not sent
> > >
> >
> > caseetc /cgi-
> > bin/caseetc/process.html PGP failed with status 3072:
> >
> > It almost seems like these processes are starting but not
completing.
> > This is the same server which was running IC 4.6, PGP and Verisign 3
> > days ago smoothly.  Now with IC 4.8 we are experiencing these
issues.
> > Any other ideas?
> >
> 
> The only thing I can think of is that if you changed the method of
> server operation, i.e. PreFork, then that could change things.
> Otherwise, I don't have a clue. We have operated Interchange 4.8 on
> very, very, busy systems without problems like this.

I made sure to stick with the same low traffic settings when we setup
the new IC.  It appears that something on my server does not agree with
the new IC 4.8.1.


> Interestingly:
> 
>     print 3072 >> 8;
> 
> is error 12, ENOMEM. Did you update Perl at the same time?

I did not update Perl or any of the modules when I installed 4.8.1.  The
server was running 4.6.1 previously.  What does the '>>' operator do?
We are running Perl 5.00503, would you recommend that we upgrade if we
attempt 4.8.2?  Any ideas why ENOMEM would be occurring when there is
plenty of memory and swap space?  It appears that this happens for any
process that IC spawns whether it is Verisign, PGP or Sendmail.

> 
> You might try PreFork mode if you are not currently using it.

In the last two days I had tried everything besides PreFork, not sure
why I didn't think of that.  Here is the list of major things I changed:

1. Downgrade from Verisign v3 to v2
2. Downgrade from PGP 6.5.8 to 6.5.2
3. Switch the IC Verisign module from Vend:Payment:Signio to
globalsub/signio
4. Install another 256mb of ram.

Last night I ended up downgrading IC from IC 4.8.1 to 4.6.1 and the
server has processed 7 orders without the problem occurring.  The only
change I needed to make for the catalog to work correctly was to remove
a nested [if][else]. The true test will be this evening when orders are
concurrent.

Since we did not see these issues in the test server I'm leery to
upgrade the server again at this point.  I'll talk with the client and
we'll figure out which route he would like to take.  He was excited to
see the speed enhancements for PreFork, but I wanted to verify the cart
worked in the same mode before changing the server level variables.

Mike let me know if you are interested in looking into this more.  This
is probably beyond my knowledge and would rather keep the client at
4.6.1 until we need a feature in the new IC.

Thanks again for your help,
-Ron