Of course the delay is due to a programmer error with the JCL
... which is a result of the fact that the operating system
doesn't offer a way to delete, if present, then allocate new in
one step, something like DISP=(RENEW,CATLG). This would have
avoided many of the unconditional IEFBR14s
The CLEANUP utility (supplied in file 183 of the CBT tape) has been
used for 20 years to delete data sets created in the rest of the job.
Sure! However, I still believe it would be a nice feature of the OS.
--
Peter Hunkeler
Credit Suisse
BPX reason code 5B1B0100
Notice: unknown modid, reason text may be incorrect
JRTrunc: Vnode operation trunc is not supported by this file system
Action: Verify that the operation was performed on a physical file
system that supports the operation.
The hint trunc seems to point to
Since you mentioned you're new to WLM, here are some quick
thoughts (some of them very simplified):
- You cannot directly set (CPU) priorities; this is WLM's duty.
- You assign work (jobs, STCs, TSO users, etc) to a WLM Service
Class (srvclass) using WLM's classification rules.
- Each service
System performance shouldn't take too big a hit from this.
But the FTP session's perfomance depends on the WLM service class
it gets assigned to. If the service class has been setup with
periods small transfers might run quickly but larger transfers
might fall through the periods with usually
Interesting, but I would think that FTP is all I/O over a
very slow path. Unless you are doing some heavy duty software
based encryption/decryption, FTP should not consume enough
CPU for WLM to even notice.
I agree. All the TCP related hints others gave are more probable
to be of help.
High Performance Java (HPJ) [snip]
It was discontinued, for a number of reasons, when Java 1.2 was
released.
Out of curiosity: What were those reasons?
I for one cannot understand what the advantage of not compiling shall
be,
especially on a platform everybody is screaming for MSU reduction.
Denis,
Thanks for the excellent argumentation. I basically concur with you.
I'd like to reply to a few arguments, though:
1. Today's JVMs offer the option to have the byte code compiled on the
fly when certain conditions are met. So, these JVMs already have the
capability to run machine code
The JIT compiler can do things that a static compiler can't. The more
frequently a method is used, the more optimisations can be applied,
such as inlining other methods and branch table reorganisation. This
results in code which can actually be faster than statically-compiled
code.
Does that
Not quite. The zAAP on zIIP feature in z/OS 1.11 applies
only if there are no zAAPs in the configuration.
zAAP on zIIP woke me up. I just got aware of this.
From the announcement letter:
z/OS V1.11 is enhanced with a new function that can enable
System z Application Assist Processor (zAAP)
Sorry to disagree but it is step initiation / allocation
that has been changed not IEFBR14.
--
Peter Hunkeler
CREDIT SUISSE
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@bama.ua.edu
Two thoughts:
- Have you tried to move the job to the library allocated
under //IEFJOBS in the master JCL?
- You might try MGCRE.
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe / signoff / archive access
IMHO, that's exactly how an intuitive application should behave.
I can only second that! TSO command processors are obliged to
implement proper ATTN handling. ISPF programs should as well.
--
Peter Hunkeler
Credit Suisse
--
We've got Tectia SSH as a replacement for telnet to login to a
z/OS UNIX shell. Unfortunately, Tectia doesn't support the chcp
command, which makes it pretty much unusable for me.
We've got an elder release (5.3.7.21) so this may well have changed.
I tried to find a hint in the doc on their web
My only question would be why has this suddenly become a
problem for jung when all other users set up with the same
script work fine?
Some possible causes:
- Are the other users sharing the uid with user START1 (probably
uid = 0)?
- Did you have automount active for /u but suddenly don't?
-
You need to be root to preserve the ownership and such.
Don't forget that you also need READ on BPX.FILEATTR.* so that
any extended attributes are preserved, too.
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe /
That plus the ease of seeing everything on one screen
without having to scroll back and forth was great.
Interesting that you can see 80 characters in the 72
positions the ISPF editor has to display the data ;-)
LRECL=72 would be the better choice, then.
--
Peter Hunkeler
Credit Suisse
Use BPXBATSL instead of BPXBATCH. It will do a local
spawn() which means your program will run as a subtask
of BPXBATSL. It will have access to all the DDs in the step.
--
Peter Hunkeler
CREDIT SUISSE
--
For IBM-MAIN subscribe /
You say you're looking at systrace entries, so I assume you've got
dump. Have you tried to follow R15 from the SVC trace entry? R15
points to the parm list. The first word pointed to by R15 is the
address of the entry point name (if the high order bit is 0).
Of course, the area pointed to by R15
(When I don't have any old newspaper or junkpaper to light a
fire for a nice barbeque, some old obsolete manuals are just
fine for that purposes...)
This gives a complete new sense to the acronym FM:
the Fire-lightning Manual
;-)
--
Peter Hunkeler
Credit Suisse
Kevin,
My assumption was that the code in question calls some function
very often. That function leads to the frequent LINKs you see.
I was further assuming that the name of the linked routine is
somewhere in that programs storage and it could still be there
at the time the dump was taken. The
What makes you think there is a batch job started?
UNIX processes (e.g. ls) inititated from another process (e.g.
your shell session) may be run in another address space. A BPXAS
address space will be started, if there currently is no idle
BPXAS that could run your command.
In this case you'll
is it related to SMF EXIT IEFUJI, our shop using IEFUJI to verify
user?
Huh? This is probably not meant as an answer to my questions I sent
you this morning, is it?
Since the questions have been sent to you instead of to the list, I'll
repeat them here:
- How do you recognize a job is
I'm sorry, but I do not understand what you're trying to find out.
First, you asked why a batch job is being submitted at login time.
Now, you talk about IEFUSI? Can you elaborate on it a bit more?
--
Peter Hunkeler
CREDIT SUISSE
I try to add OMVS subsystem with SETSSI command it looks successfully
There is no need to add this SSI entry. OMVS is as much a subsystem as
STC and TSO are: None. These terms appear in the IEFUSI parm list so
that you can decide what kind of address space your dealing with.
Why dou you think
In the JCL below, what tells the BPXBATCH program to
execute the commands in the file specified on the STDENV DD ?
I would be astonished if anything from STDENV would be executed.
STDENV is a DD statement read by BPXBATCH *before* starting
whatever it has to start according to the PARM=.
Every
What he said, but I would do both
/*
//
While a // definitely denotes the end of the JCL for the
current job, /* does not. /* terminates the current sysin
dataset (note that DLM= may denote a different sequence).
Why not put the data on the stack and use
address TSO submit *
from the REXX
Have you checked your primary selections,
additional selections as well as set mask character
in the filter menu?
The latter fooled be recently since I'm used to use *
as the generic character whereas EJES defaults to .
--
Peter Hunkeler
CREDIT SUISSE AG
Well, well. I amused about the discussion in this thread (no
offence intended). So many posts, but still neither the actual
abend including the reason, nor the PSW, nor the failing
instruction including the analysis what its operands are have
been posted.
Neither has proof been provided that the
WWwsswWW
And this means what?
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO
Search the archives at
The A/M attibute is a dataset attribute that is mostly of
importance to programs preparing the data to be printed, i.e.
JES2/3 or PSF/Infoprint. JES2/3 needs this information to build
the channel programm to drive the printer. PSF/Infoprint needs
this information to format the data according to
Interesting how different this is from below-the-line...
In what respect do you think it is different?
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to
Tommy,
To quote the Computerworld article:
IBM has expanded its server lineup with a new mainframe system designed just
for
Linux that may be aimed, in particular, at higher-end x86 systems.
The keyword is expanded.
And in fact it is nothing new, IIRC. You could have IFL only machines for
You need to talk to your SMS colleagues. If the ACS routine assigns
a STORCLAS, then your data set will be SMS managed and there is
nothing you can do about it.
ACS routines usually decide on dsname and possibly on some other
attributes whether to put this data set under SMS control. Your
SMS
And now watch the guilty parties spin it as we need to
migrate off these mainframes to HP hardware...
.. and in support of their ignorance they surely will refer to
some glamorous consultant reports like those that recently
recommended ... now being the time to get ... (to somewhere
else :-)
It improves performance by avoiding the extra level of indirection
when translating the virtual address to a real address on a TLB
miss.
Doesn't have to reference the region-table entry.
That would imply an architecture change. The segment table index
has long been 11 bits which corresponds
The last .boo version of the POO is SA22-7832-03.
... which has the nice side effect that I no need to download a 32MB
file before I can start a search :-((
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe / signoff /
Hi,
I'm about to update a chart I'm using to illustrate the
relative speed of various actions like CP cycle, storage
access, etc.
Does it still hold true that an I/O from controller cache (cache
hit) takes about 1ms and and I/O from the platter (cache miss)
takes about 5ms. All I want is two
Maybe not officially, but on IBMLink, for APAR OA25204.
Why can't we bury this stupid discussion once for all?
--
Peter Hunkeler
CREDIT SUISSE AG
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email
As a follow-up to last weeks discussion, I stumbled across a
short document I wrote in 2006 when I was researching how
this works. If anybody is interested, just drop me a note.
BTW, it also contains are sample asm routines to try yourself.
--
Peter Hunkeler
CREDIT SUISSE AG
If you want to stay subscribed but don't want individual
mails sent, send a mail with:
set ibm-main nomail
to listserv at bama dot ua dot edu (change this to a valid
email address by replacing the at with @ and the dot with
a .).
If you would like to get a single summary type mail per day send
Unfortunately, NO.
Thit is not a correct answer, see below.
I went through the same thing, when we first implemented UNIX
System Services. This is z/OS's (previously ESA 5.2 OS/390) method
of emulating the UNIX fork of a process.
A spawn is even 'worse', since the originating address space
Change 'spawn' 'exec' ALL ?
No. exec() translates (more or lesss) to next job step.
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@bama.ua.edu with the
But, there was/is a case with ESA 5.2, where there is only
the forked/spawned/whatever USS piece is the only part left,
and the originating/calling address space is no longer with
us. It's a USS process, with no apparent parrent, and caused
us grief when we were trying to 'get it right' with
I refused to design a WLM policy to address anomalous behavior.
z/OS UNIX was nothing but anomalous behaviour in the eyes of z/OS MVS,
but z/OS UNIX (MVS OpenEdition with maiden-name) married-in and they
finally managed to respect and support each other ;-)
--
Peter Hunkeler
Credit Suisse
It's been a while...but this is what I seem to remember
(too lazy to RTFM right now):
ftp's children run under the userid with which the
remote user logged in; and the jobname is derived from that
userid as well. You can't catch ftpd's child processes
with ftpd's userid.
Subsystem Type . : OMVS
Please excuse me for sending that incomplete post. I had the intention
to save instead of send right know because I wanted to first confirm
a statement I was about to make
--
Peter Hunkeler
CREDIT SUISSE AG
--
For IBM-MAIN
Nevertheless, the rule you posted is assigning SYSSTC to each an
every file transfer.
No, only the FTP daemon. I think you are assuming I am using
_BPX_JOBNAME because I mentioned it. We don't.
My bad, yes I was assuming you're using _BPX_JOBNAME for the ftpd.
Are you saying I can do
I'm looking for some direction and advice on dynamically extracting
STDOUT data from USS that I can feed into a CICS sub-system. Would a
writer / listener be the right direction ?. How would I extract the
STDOUT data ?.
Create a named pipe and have the writer send its stdout to that pipe.
If the OP actually wants to write to spool, and is not necessarily
running under BPXBATCH, he can do:
I can't see where the OP ever said he wanted to write to the spool.
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe
We have done some tests and found that there is no issue to read
and write HFS files from a CICS transactions, using the C library
functions for reading and writing files.
I'm by far anything else than a CICS expert, but I seem to remember
that CICS transaction must not go into a wait, be it
Gee, you do I/O, you wait ... [snip]
Sure. I don't know if there is a CICS service to do I/O
to some non-CICS dataset, or if it was simply tolerated.
Actually, a wait only affects other transactions in the
same CICS region as the will have to wait, too. I remember
that in the early releases of
What happens when a CICS transaction must do something
like:
Invalid PIN number; please re-enter
??? Does that spawn a separate transaction?
The basic design of transaction managers is that there
is a message queue manager which stacks messages until
a transaction becomes idle so it can
Isn't it the other way 'round?
The operator tells the printer to print output matching
ceratin criteria, such as FORMS ($tprtxyz FORMS=FORM1). This
effectively tells JES to only reply to PSF's is-there
anything-I-can-print-inquiries with output that has
FORMS=FORM1. When there is no more output
The confident assertion of radically ignorant conjectures
is not so defensible. Do your homework before you post.
May I hope that you'd offer to act as censor so we can avoid
having to read radically ignorant conjectures in the futurre?
--
Peter Hunkeler
Credit Suisse
The C runtime library has only the exec-family of functions. The
documentation says they require the program that is to be started
to reside in the HFS.
You might try with exec()inf to an external link. The external link
is an HFS entry that points to the MVS search order.
o ln -e MVSPGM
That does not solve all my problems, though.
What makes you believe BPX1EXM would be available to
non-UNIX users? My guess is, it is not. Why? Well, ...
... it starts with BPX, which denotes it as belonging to
the z/OS UNIX component.
... it is described in the z/OS UNIX Assembler Services
I have read that these services are part of the OS, and always
available, whether or not the user has access to OMVS / HFS or not.
Where did you read that? Sure, all BPX services are part of the OS,
as are all the other services z/OS offers. Some, however, have some
prerequisites in order to
That seems like an acceptable requirement for us to put on
our customers, doesn't it?
It sure is in my opinion. Also, I have always told people to
think twice before activating the default user thing and
to consider assigning uid/gid to each and every MVS userid,
at least to those that are
o And those that are allowed to submit batch jobs.
o And those that are allowed to use FTP.
o And, of course, those that are allowed to log in to a shell.
I didn't want to go as far as saying z/OS UNIX is ubiquitous,
and therefore any RACF userid should have an OMVS segment
(although I
I've just found an interesting new ability in z/OS 1.11.
If you define the BPX.UNIQUE.USER in the FACILITY class,
then RACF itself will find an unused UID/GID and update
the user to have an OMVS segment.
I for one would rather not have arbitrary uids and gids.
But if you don't care then this
What makes you believe BPX1EXM would be available to
non-UNIX users?
What is a non-UNIX user? No OMVS segment? Not yet dubbed?
None of the above?
It is perfectly clear from the context what was meant.
--
Peter Hunkeler
Credit Suisse
I got absolutely NO diagnostic information. I'm totally
frustrated and ready to go postal.
No hint in LOGREC?
-
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to
2. It depends on datasets activity. In fact catalog is usually used
at file open and close.
I may well display my ignorace, anyway: It is my understanding that
the catalog is only accessed at allocation time to find the volume
that data set resides on. VSAM data sets being an exeption
4) OMVS task is exempt from JWT
What exaclty do you mean by OMVS task?
o the kernel (STC name = OMVS)?
o a TSO user that called OMVS to start a shell?
o rlogin/telnet/ssh users logged into a shell?
o a program having called any z/OS UNIX service and
which is therefore dubbed a process?
o
The length in the length field in the structure field introducer must
match the number of bytes in the record, including this length field,
excluding the X'5A' byte. The length of the record (RDW) is this
length plus 1 for the X'5A'.
So, for the VB format, the length must be consistent. Adjust
Basically, my problem was that I had an invoke data map command
where the name of the data map itself was less than 8 characters.
With fixed length format the record looked like this:
5A0010D3ABCA00D7F2C4E2D4E3F140404040404040404040... (with
spaces (x40) padding to the end of the record)
Hey zMan,
I suggest you identify yourself. It's just part of the netiquette (at
least on this list) not to stay anonymous.
That especially applies when you're ranting...
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN
I've beein disliking zAAPs ans zIIPs from the beginning. Not
only require they a more complex dispatching algorithm but ---
having read the below mentioned paper --- it becomes clear that
they also may put system intergrity at risk.
In order to lower software cost, programmers are encouraged
to
the command has it is on AIX :
df -k | grep -e '9[0-9]%' -e 100%
I just tried:
df -P | grep -e '9[0-9]%' -e 100%
Seems to do just what you want. Using putty and the sh shell.
Do you get any errors? What do you mean with It refuses to work?
--
Peter Hunkeler
Credit Suisse
[snip]
QUEUE,
$E
SUBMIT * END($E)
As an alternative, simply queue an emtpy element ( queue ) and SUBMIT
will stop reading w/o the need for the END($E)
--
Peter Hunkeler
Credit Suisse
An SRB receives control in supervisor state. And while it can L(E)PSW
itself to problem state, it will not be able to get back to supervisor.
The manual says:
9.5.1 Implications of Running in SRB Mode
The SRB routine runs in the operating mode known as SRB mode.
Code in SRB
If your emulator is setup to run with CP037, then you need
to do one of the following:
a) tell the shell to expect characters in CP037:
export LC_ALL=En_US.IBM-037
then restart the shell
exec sh -L
b) tell the OMVS TSO command processor to translate characters
as they are
I admit I'm haven't completely understood what you want
to do, but
o Why bother writing a listener by yourself if you can
take advantage of inetd. From what I understood, it
does what you need: Start a new process for each new
login.
o Why do you need another fork() of your login
IEBCOPY does not need APF authorization for every
operation. Is step 8 doing something different from all
the other steps? Something that might need APF auth.?
--
Peter Hunkeler
CREDIT SUISSE AG
--
For IBM-MAIN subscribe /
IEBCOPY does not need APF authorization for every
operation. Is step 8 doing something different from all
the other steps? Something that might need APF auth.?
AFAIK the above is not exact. IEBCOPY does check APF authorization
before run. Try the following: copy IEBCOPY member to your
... When the job was restarted from the failing step, no change
was made to the JOBLIB or the Step itself and it ran successfully
Out of curiosity, and since I just learned that IEBCOPY will
issue message IEB1099I at startup if it finds itself not running APF
authorized: Would you have a look
Because I believe that some reactionary shops don't like z/OS
UNIX System Services and I'm considering pandering to them.
It's not a question if you (sorry, I mean they) like it or not.
If they run z/OS, they have got UNIX. They'd better be prepared
for the next update of one of their IBM or
This is far simplier than a RACROUTE REQUEST=VERIFY or
IRRSIA00 in that it does not require supervisor state or
APF authorization, but the JrEnvDirty seems to imply the
program must be program controlled. I know what that
means for PDS resident programs, but not HFS resident
programs. At
This is not only a problem of universites not teaching
its students COBOL, PL/I, IBM Mainframes, etc. It is also
a home made problem of the companies requiring this
kind of skills. Many of them had their own IT school
with which they took care of educating employees in the
skills they need for
Found the problem
If go into ISPF/EDIT on the ds it chops off trailing blanks
PRESERVE ON it a good idea, except when you really want to
truncate trailing blanks.
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe /
But beware; PRESERVE may not do what you intend...
Thanks for remembering me of the real function of that
command: preserving the LRECL (and not preserving trailing
blanks).
I only ever needed to change data without changing the
record's length recently. I guess that's why the real
meaning
[snip]A GENER of the files to VB prior to sending them to
OMVS indeed confirmed this is what was happening.
Anyone else seen this - or got an explanation ?. gil maybe ... ?
How did you move the OFILE to the filesystem?
--
Peter Hunkeler
CREDIT SUISSE AG
cp //'..' /tmp/file.txt
and/or cat //'..' /tmp/file.txt
Before posting my question, I ran a LISTC with OFILE, verified
it is VBA-140, then did
cat '//TEMP.LISTCOUT' temp.listcout
I'm baffled that I don't see the ASA char when doing a
cat temp.listcout
afterwards. anyway, I don't see
[me]... Will have to think about it a bit more.
I think the explanation can be found in the C/C++ Programming
Guide, in the chapter about Using ASA text files (Chapter 7
in the recent editions).
Thanks a lot, Bill, for the nice summary!
--
Peter Hunkeler
Credit Suisse
Creating the data set as RECFM=V on the z/OS side and omitting
any filtering downstream seems optimum to me.
What arguments favor a more complicated process?
The fact that the program may not support writing variable length
records?
--
Peter Hunkeler
Credit Suisse
I remember much levity in my time at Amdahl and IBM's inability
to build a machine that scaled past 10 (?) engines.
We managed to skip past that o.k. - didn't have anything to do
with limitations in the OS - all the relevant control blocks had
plenty of width ...
Given the current powerPC
Creating the data set as RECFM=V on the z/OS side and omitting
any filtering downstream seems optimum to me. What arguments
favor a more complicated process?
I'll try it.
It didn't read right. I'll just clean up the file on the Unix side.
So then the program did not write variable length
I would kindly ask IBM to change the description in a way to be
compatible to the ANSI standard and to that what the compilers,
in fact, do.
You need to send your doc change request as an RCF. Send an eMail
with your comments to mhvr...@us.ibm.com
I have made good experience, although it may
Why is text in IBM-1047 more stable that text in ISO8859-1?
ISO8859-1 are basically ANSI characters.
(ANSI -- antsy, just in case it is just too weird for some).
I know, if I have to explain it, it isn't really funny.
Not necessarily! If you're not a native speaker, like me,
you might simply
o Kudos to FTP server -- the messages are excellent.
Could only be topped if the message would the who the
holder was.
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send
Would anyone care to submit a Requirement that ISPF defer
requesting the EXCL SPFEDIT until the first SAVE?
I would vote against it. Definitely! If I'm in EDIT, I expect
that nothing can change the content under the covers, no matter
whether I'm going to actuially change the content or not.
I don't have access to a uid=0 userid, so I cannot
test myself. Anayway,
... what are the differences in the OMVS segments of
both users?
... can the uid=0 userid edit plain standard MVS datasets?
... any messages in any of the logs (JESYSMSG, ISPLOG, SYSLOG)
--
Peter Hunkeler
CREDIT
Ed's example concerned retrieving the member, not storing it.
It seems this is not my best day for posting. I'll go look
for a new pair of glasses; These seem to blank out words :-)
--
Peter Hunkeler
Credit Suisse
--
For
Could only be topped if the message would the who the
holder was.
It does. Re-quoting from the OP:
is held by: 008F EDJXADM EXCL on SPFEDIT
My appologies go to FTP!
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN
Cool. And it turns out ASCBASID isn't really for the home address
space,
ASCBASID is an attribute of the examined ASCB. ASCB's are in common
and the ASCB for any address space can be examined by any address
space.
...and any address space can be HOME, PRIMARY or SECONDARY at the
same time,
Is your system (i.e. the hardware) set to local time or GMT?
What does a D T from the MVS console display?
--
Peter Hunkeler
CREDIT SUISSE AG
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to
Not very intuitive, I admit, but that's my interpretation
of how it is supposed to work.
non-serious
I thought that only the designers of COND= were drunken,
but is seems those desinging the IF/THEN/ELSE/ENDIF may
have been, too.
/non-serious
--
Peter Hunkeler
Credit Suisse
I will continue to pursue documentation to see what might be
incorrect or missing.
The setting from /etc/init.options (not init.profile) is only
used at IPL time when BPXOINIT initializes. Part of that
initialization is to read /etc/init.options, and run the script
specified via -sc (usually
I've got a JES2 background, and I seem to remember that with
JES2 one of the printer setup parameter is the routcode. Local
destinations can be assigned to the routcode and JES2 would then
select output matching these destinations for printing on
these decives.
From reading JES3 manuals, I've
1 - 100 of 214 matches
Mail list logo