INFO-VAX Wed, 16 Jul 2008 Volume 2008 : Issue 394
Contents:
Re: DCL divide by zero WAS: Symbol Substitution Mystery
Re: DCL divide by zero WAS: Symbol Substitution Mystery
Re: Monitoring Runaway CPU Processes
Re: Monitoring Runaway CPU Processes
Re: Monitoring Runaway CPU Processes
Re: Q: Configure TCP/IP Services for a node not yet booted?
Re: Q: Configure TCP/IP Services for a node not yet booted?
Re: Q: Configure TCP/IP Services for a node not yet booted?
TCPIP smtp queue - how to change it?
Re: TCPIP smtp queue - how to change it?
Re: TCPIP smtp queue - how to change it?
----------------------------------------------------------------------
Date: Tue, 15 Jul 2008 18:22:40 -0700 (PDT)
From: AEF <spamsink2001@yahoo.com>
Subject: Re: DCL divide by zero WAS: Symbol Substitution Mystery
Message-ID: <1381471f-1faa-4973-a138-7c2d553ed762@d77g2000hsb.googlegroups.com>
On Jul 14, 8:52 am, koeh...@eisner.nospam.encompasserve.org (Bob
Koehler) wrote:
> In article <2c5155ac-72ca-423c-915c-76be0df6b...@r66g2000hsg.googlegroups.com>, AEF <spamsink2...@yahoo.com> writes:
>
>
>
> > I don't have one and there's no way my company is going to get one,
> > especially just for this. Fortunately I almost never need to divide on
> > the VAX. The only example that comes to mind is the calculation of
> > disk space percentages.
>
> If you have VMS, you have Macro-32, either as an assembler on the
> VAX or as a compiler on the other two.
Well, I'm not going to learn Macro-32 just to calculate disk space
percentages. I don't have the time, anyway.
While we're on the topic: Suppose I wanted to port my trading app from
VAX to Alpha or Integrity. All of it is written in DCL and Pascal
except for the "messenger" program, which is written in Macro. What
would it take to move this macro code to Alpha or Integrity (and why
is it called "Macro")?
This is what it does.
;Summary: This program functions as an agent for task-to-
task
; communication messages among participated
processes.
; Program is also responsible for maintaining
message
; queue for each process in an asynchronous
environment.
Thanks!
AEF
------------------------------
Date: Tue, 15 Jul 2008 21:31:08 -0400
From: "Richard B. Gilbert" <rgilbert88@comcast.net>
Subject: Re: DCL divide by zero WAS: Symbol Substitution Mystery
Message-ID: <c6SdncwZjIyQzeDVnZ2dnUVZ_tfinZ2d@comcast.com>
AEF wrote:
> On Jul 14, 8:52 am, koeh...@eisner.nospam.encompasserve.org (Bob
> Koehler) wrote:
>> In article <2c5155ac-72ca-423c-915c-76be0df6b...@r66g2000hsg.googlegroups.com>, AEF <spamsink2...@yahoo.com> writes:
>>
>>
>>
>>> I don't have one and there's no way my company is going to get one,
>>> especially just for this. Fortunately I almost never need to divide on
>>> the VAX. The only example that comes to mind is the calculation of
>>> disk space percentages.
>> If you have VMS, you have Macro-32, either as an assembler on the
>> VAX or as a compiler on the other two.
>
> Well, I'm not going to learn Macro-32 just to calculate disk space
> percentages. I don't have the time, anyway.
>
> While we're on the topic: Suppose I wanted to port my trading app from
> VAX to Alpha or Integrity. All of it is written in DCL and Pascal
> except for the "messenger" program, which is written in Macro. What
> would it take to move this macro code to Alpha or Integrity (and why
> is it called "Macro")?
<snip>
VAX to Alpha is a piece of cake. Recompile the Pascal, "recompile" the
Macro, link and run it. The only potential problem that I can recall is
code that depends on a 512 byte page size. There is documentation
somewhere on VAX to Alpha porting.
I've never even seen an "Integrity". You'll have to ask someone who
knows. There may be some documentation on porting VAX to Integrity,
google for it.
------------------------------
Date: 15 Jul 2008 18:06:47 GMT
From: VAXman- @SendSpamHere.ORG
Subject: Re: Monitoring Runaway CPU Processes
Message-ID: <487ce737$0$7342$607ed4bc@cv.net>
In article <YcmdnYeKz9EWUOHVnZ2dnUVZ_h7inZ2d@comcast.com>, "Richard B. Gilbert" <rgilbert88@comcast.net> writes:
>James J. O'Shea wrote:
>> Hello,
>>
>> Does anyone know if there is a utility, free or
>> otherwise, that will alert when there is a runaway
>> CPU process?
>>
>> Thanks,
>> Jim O'Shea
>> Chicago, IL USA
>
>Define "runaway process".
I was going to ask the same question, i.e. what is a runaway process?
If you are computing PI to the gazillionth decimal, you're likely to
see a huge run-up in CPU utilization akin to a process in an infinite
loop or, my favorite, a process with the no delete bit set and an un-
handled exception has ocurred. Which process should be killed off or
set off an alarm, etc?
>If you are talking about an "infinite loop", you can shut it down with a
>limit on CPU time; e.g. RUN /TIME_LIMIT=. There may be quotas and
>limits you can set in AUTHORIZE that will shut it down for you.
>
>You can avoid the whole problem by not coding infinite loops!
No more:
while (1)
{ code }
Instead:
for (i=0;i<0x7FFFFFFF;i++)
for (j=0;j<0x7FFFFFFF;j++)
for (k=0;k<0x7FFFFFFF;k++)
for (l=0;l<0x7FFFFFFF;l++)
for (m=0;m<0x7FFFFFFF;m++)
for (n=0;n<0x7FFFFFFF;n++)
{ code }
:)
--
VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)COM
"Well my son, life is like a beanstalk, isn't it?"
Copyright 2008 Brian Schenkenberger. Any publication of _this_ usenet article
outside of usenet _must_ include its contents _in_its_entirety_ including this
copyright notice, disclaimer and quotations.
The citizens of our state must be free, within reason, to speak out on matters
of public concern. So long as they state the facts implicated fairly and
express their opinions, even in the most colorful and hyperbolic terms, their
speech should be protected by us. -- NJ Superior Court Appellate Div. (NJSC)
... pejorative statements of opinion are entitled to constitutional protection
no matter how extreme, vituperous, or vigorously expressed they may be. (NJSC)
"Coding is _not_ a crime!" Support the EFF: http://www.eff.org
------------------------------
Date: Tue, 15 Jul 2008 13:12:57 -0700 (PDT)
From: Ed Wilts <ewilts@ewilts.org>
Subject: Re: Monitoring Runaway CPU Processes
Message-ID: <0f4fabc1-8272-4fd1-816d-eef8b591912d@a70g2000hsh.googlegroups.com>
On Jul 15, 11:02=A0am, "James J. O'Shea" <seamas_...@ameritech.net>
wrote:
> Does anyone know if there is a utility, free or
> otherwise, =A0that will alert when there is a runaway
> CPU process?
There are ton, on the freeware CDs, and otherwise on the net. The
hard part, as others have mentioned, is defining a runaway, and teh
problem is a *lot* harder than it looks, especially if you want to
factor in I/O. For my system, I use a very, very low-tech solution -
a simple DCL script that monitors the total CPUTIM of a process. If
it's a certain image, it gets skipped, since I know it uses a lot of
CPU. If it's other known images, I automatically kill them. For the
remainder, I mail the sysadmins and let them/us figure it out. If you
have a stable environment with the rare bad process, this can be very
effective. In other environments, this approach can suck.
.../Ed
------------------------------
Date: Tue, 15 Jul 2008 16:48:11 -0400
From: JF Mezei <jfmezei.spamnot@vaxination.ca>
Subject: Re: Monitoring Runaway CPU Processes
Message-ID: <487d0d28$0$18522$c3e8da3@news.astraweb.com>
Ed Wilts wrote:
> problem is a *lot* harder than it looks, especially if you want to
> factor in I/O.
But one must be careful with that with newer apps (by VMS standards).
Mozilla for instance will eventually garner all system resources but
will get both the DECWSERVER and itself running CPU bound, as well as
both generating tons and tons of hard page faults once either of them
has reached the page file quota.
From a "reality" point of view, it has become a runaway process. But
from a pure CPU perspective it doesn't appear as such since:
-work is split between 2 processes
-there is plenty of IO happening
This is caused by Mozilla's infamous memory leaks which also result in
Mozilla forgetting to tell the X server it can deallocate pixmaps.
------------------------------
Date: Tue, 15 Jul 2008 16:30:46 -0700 (PDT)
From: Ken.Fairfield@gmail.com
Subject: Re: Q: Configure TCP/IP Services for a node not yet booted?
Message-ID: <723f553f-fce6-4541-ab1c-48f5a0c05068@2g2000hsn.googlegroups.com>
On Jul 15, 3:01 am, JF Mezei <jfmezei.spam...@vaxination.ca> wrote:
[...]
> Note to the OP: Make sure you scan thorugh TCPIP$CONFIG.COM to look at
> the code that determines, based on name of the ethernet device, the name
> of the interfaces you will have on that system.
>
> If the ethernet device names on all 4 systems are the same device names,
> then you can assume the same interface names. But if the ethernet
> devices are different, then you'll have different interface names.
>
> here is the relevant definition:
> TCPIP$EDEV = "0 XE:DE XQ:QE ES:SE ET:NE EX:XE EF:FE EZ:ZE EC:CE" + -
> "ER:RE EW:WE EB:BE EI:IE LL:LE EG:GE VL:VE"
>
> My ethernet devide is an EWA0, which means my internet interface is WE0
> (from the EW:WE item)
These systems have either 4 or 5 devices each, but all are EW's.
So node D has EWA0, EWB0, EWC0 and EWD0, but node C
has in addition EWF0. So I have WE0-WE3 on two nodes and
WE0-WE4 on the other two nodes.
That does mean that any interface configuration must be done
from the node that "owns" the interfaces.
Still, my configuration is simple enough: all services that can
be enable cluster wide are; the SYS$SPECIFIC directories are
only TCPIP$NTP, TCPIP$BIND, TCPIP$SMTP, and TCPIP$ETC,
the later of which has only template files (i.e., .DAT files with
comments).
So it looks like there are the 7 logicals pointing to "database"
files in SYS$COMMON:[SYSEXE], and if I define those
appropriately, e.g., on node C with DSA10: mounted, I'd
think I could run TCPIP$UCP.EXE from DSA10: and "see"
all the configurations...and add those for node C.
I would also expect the directories for BIND and NTP,
etc., to get created on DSA0: in C's SYS$SPECIFIC,
but those can be copied to DSA10 after the fact.
So what is the gotcha here? (I intend to avoid
TCPIP$CONFIG.COM for a variety of reasons.)
Thanks, Ken
------------------------------
Date: Tue, 15 Jul 2008 16:33:44 -0700 (PDT)
From: Ken.Fairfield@gmail.com
Subject: Re: Q: Configure TCP/IP Services for a node not yet booted?
Message-ID: <195864e8-e22b-4657-b60b-57876050c011@z72g2000hsb.googlegroups.com>
On Jul 15, 3:29 am, "R.A.Omond" <Roy.Om...@BlueBubble.UK.Com> wrote:
[...]
> To Ken (Fairfield), I was just about to add a similar reply to JF's.
>
> Unless the interfaces are all the same on all 4 systems, you're going
> to get into a *huge* mess, even if you manage to get round the
> indexed-by-node-name issue.
I'm not sure which one of JF's creative solutions you
are thinking of, but if I configure a given node's interfaces
while logged into that node, how do I get into a mess
with indexed-by-node-name issues???
> I think you'd be well advised to not go ahead with your original plan.
Thanks for your concern, Ken
------------------------------
Date: Tue, 15 Jul 2008 16:37:15 -0700 (PDT)
From: Ken.Fairfield@gmail.com
Subject: Re: Q: Configure TCP/IP Services for a node not yet booted?
Message-ID: <17007f40-e287-4a29-a910-648cbe948721@x41g2000hsb.googlegroups.com>
On Jul 15, 9:14 am, JF Mezei <jfmezei.spam...@vaxination.ca> wrote:
> R.A.Omond wrote:
> > Unless the interfaces are all the same on all 4 systems, you're going
> > to get into a *huge* mess, even if you manage to get round the
> > indexed-by-node-name issue.
>
> Not necessarily. The logic to determine interface name is pretty
> straighforward in tcpip$config.com
>
> Once you have Node D all configured and running, it makes for a good
> template to see how to do the other nodes manually.
Why do I need to know anything to worry about that logic,
or indeed, why do I need to think of node D's configuration
as a template, if I'm do node C's configuration from node C?
As I've said more than once in this thread, I intend to avoid
TCPIP$COM.COM which likely *does* make assumptions
about the product being installed on the system disk that
is the node's boot disk.
-Ken
------------------------------
Date: Tue, 15 Jul 2008 13:47:40 -0700 (PDT)
From: ythefoe@gmail.com
Subject: TCPIP smtp queue - how to change it?
Message-ID: <c68e84a8-0ffb-409d-ab7b-b48db6d47d80@l64g2000hse.googlegroups.com>
Hello,
When you do a 'TCPIP SHOW CONFIGURATION SMTP', you'll see at the
bottom line the associated SMTP-queue.
My question is, how can you change this SMTP-queue.
I found in the documentation that you can set the number of queues
(tcpip set configuration smtp/queue=n), but I can't find nothing
telling me how to change it...
I did not find any configuration file; logical,... where this queue is
kept...
Another question related to this: these SMTP-queues, are these normal
queues (created with initialize)?
Thx in advance....
Me
------------------------------
Date: Tue, 15 Jul 2008 17:13:46 -0400
From: JF Mezei <jfmezei.spamnot@vaxination.ca>
Subject: Re: TCPIP smtp queue - how to change it?
Message-ID: <487d1326$0$18587$c3e8da3@news.astraweb.com>
ythefoe@gmail.com wrote:
> (tcpip set configuration smtp/queue=n), but I can't find nothing
> telling me how to change it...
SET CONFIG stores those changes in the TCPIP$CONFIGURATION.DAT file.
When the mail software starts, it looks at its records to pickup the
config. Not sur who gets to create the new queues when you enlarge the
number though.
> Another question related to this: these SMTP-queues, are these normal
> queues (created with initialize)?
Yep those are just queues. Instead of a print symbiont, theyr have a
different symbiont that processes mail entries. Mail entries consist of
2 files. A proprietary undocumented control file, and the actual
contents of the email RFC922 header + text.
It s the proprietary control file which is "submitted" to the SMTP queue.
------------------------------
Date: Tue, 15 Jul 2008 22:21:20 GMT
From: =?ISO-8859-1?Q?Jan-Erik_S=F6derholm?=
Subject: Re: TCPIP smtp queue - how to change it?
Message-ID: <Ap9fk.1010$U5.681@newsb.telia.net>
ythefoe@gmail.com wrote:
> Hello,
>
> When you do a 'TCPIP SHOW CONFIGURATION SMTP', you'll see at the
> bottom line the associated SMTP-queue.
> My question is, how can you change this SMTP-queue.
Why would you "change" (whatever that is) the SMTP queue(s) ?
Jan-Erik.
> I found in the documentation that you can set the number of queues
> (tcpip set configuration smtp/queue=n), but I can't find nothing
> telling me how to change it...
> I did not find any configuration file; logical,... where this queue is
> kept...
>
> Another question related to this: these SMTP-queues, are these normal
> queues (created with initialize)?
>
>
> Thx in advance....
> Me
------------------------------
End of INFO-VAX 2008.394
************************