🚀 go-pugleaf

     Hello all.  Happy New Year!

     I have what I believe to be performances problem.  Curiously, I
only see them on one disk in my entire VAXCluster.  It happens to be the
disk which holds PMDF and JNET.  I do not believe either of the systems
to be directly responsible for the problem.

     It may be a bit hard to explain because it's the feel of the system
when sending e-mail via PMDF that shows the problems.

     I have PMDF 4.1 set up to run on my VAXCluster running VMS 5.5 on two
6420's with TGV MULTINET 3.6.  I've see this same problem on other
configurations of VMS, hardware, Wollongong rather that TGV, so it's not
specific to the current arrangement.

     I have PMDF defined on all the systems but only one is running the
transport layer SMTP etc. programs.  This system pumps anything from
1000 to say 8000 message as day or more.  As long as everything works
life is fine.  Now and then at least one of two things happens;

     a) exiting from typing a mail message sent via PMDF takes many
        minutes, people have complained of up to 15.  Processes are
        stuck in LEF waiting for disk apparently.

     b) the outgoing mail que backs up because several other sites
        are very slow at responding with either TCP/IP connects or
        SMTP dialogue.

     The first one seems to relate to how many files are on the disk.
When this long pause happens I find I can clear it by moving things
around to other disks.  DEC hasn't been very forthcoming with any
definitive explanation.  Seeing that there are LOTS of small files
coming and going all the time I think maybe RMS is getting confused by
extreme fragmentation.  We run Rabbit 7 (Perfect Disk now I think) but
it never has a chance to do a good job.  Other than find a BIG empty
disk to house mail (expensive) I'm not sure what to do with this one
yet.

     The second problem is additive.  Since I only have one outgoing que
the affect of several slow receivers clogs my que.  I'm running a
VAXCluster here.  I have two nodes (three if you count a 3100 in the
corner).  I know that I can run local delivery on both nodes.  I've
never seen anything in the manual about running outgoing Internet mail
on multiple nodes.  I don't remember seeing it in the manual but I have
opened up the one running batch que to run more than one deliver at a
time.  This seems to work although the process name set fails.  But the
mail gets delivered.

     Would it help to run another que on the second node?  Some mailers
do reverse lookup via PTR records.  PMDF is set up to announce the
primary node as the source in envelopes so a reverse lookup will fail. I
don't know what would happen in this case.

     So, any thoughts about either problem?

     Thanks much.

Chris Johnson
Northeastern University