Re: ./ ADD - which utility? (RANT)
On 21/05/2024 12:50 am, Radoslaw Skorupka wrote: The only thing I still cannot do is to convince my students the z/OS<->PC data exchange is easy and well tooled. It will never be easy due to complications from EBCDIC and record structure. Even IBM frequently gets it wrong. There are new tools, e.g. I have been experimenting with Zowe CLI. It's worth trying for file transfer and can create new datasets, PDS etc. It only requires z/OSMF on the mainframe, i.e. nothing to install there - only configuration. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Mainframe performance tool replacement
On 8/05/2024 7:40 am, Wayne Bickerdike wrote: EasySMF. Andrew Rowley may chip in here. Thanks, yes EasySMF is another option. There is a Windows based component with canned reports, and Java SMF API for customized reports that can run on z/OS. I'm happy to work with you to reproduce some of your current reports with a free trial of the Java API, if that interests you. More detail is available here: https://www.blackhillsoftware.com/ Andrew Rowley -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Anyone exploiting ZEDC?
On 19/04/2024 9:35 pm, Scott Chapman wrote: Agreed, it's been many years since I ran similar test with our software and don't remember the exact details, but my recollection is that I saw no noticeable change in CPU consumption. Which I remember being particularly interested in because I'd heard of issues reading that compressed data. I did some experiments with SMF data with and without zEDC. What I saw in my test: - zEDC I/O was much faster, reducing elapsed time by more than 10x in some cases - CPU time for compression seemed to be less than 0.2s per GB - CICS software compression greatly reduced the CPU time for copying the zEDC compressed data e.g. if you have weekly/monthly SMF housekeeping Also some interesting results processing zEDC compressed data with Java. zEDC actually reduced the amount of time on the general purpose CPs - it seems like more of the work ended up on the zIIPs. Reporting on 39GB of zEDC compressed CICS data used only 2s CP / 38.6s zIIP time, compared to 5.6s CP time reading the data with IFASMFDP. Reading the 6.6 GB dataset in Java used 1.49s CP time without zEDC, 1.25s CP time with zEDC (plus zIIP time). You can read the details here: https://www.blackhillsoftware.com/news/2024/04/29/using-zedc-compression-for-smf-data/ -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: REXX vs other languages WAS: Rexx numeric digits and scientific notation question
On 23/04/2024 11:55 am, Paul Gilmartin wrote: On Tue, 23 Apr 2024 10:59:47 +1000, Andrew Rowley wrote: ... To me, it is much clearer to be explicit, including the concatenation, e.g. "DELETE " || foo That overkill is apt to confuse a POSIX shell partisan who would see the blank as part of the command name Isn't Rexx assembling it all into one string, which gets then run as the command? There can be significant spaces between parts of the command, and no spaces allowed in other parts. So you could have e.g. "DELETE ABC" || suffix or "DELETE" prefix || suffix which is very different from "DELETE" prefix suffix and it can get much more complex. I would rather use a consistent concatenation operator and explicitly insert spaces rather than rely on getting the right concatenation operator in the right place. It's a matter of style, but there are some things that just look like potential bugs to me. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: REXX vs other languages WAS: Rexx numeric digits and scientific notation question
On 23/04/2024 12:01 am, Seymour J Metz wrote: delete foo rather than 'DELETE' foo is a simple example. In practise I often need multiple upper case constants in a single expression. In this example, delete *looks like* an instruction rather than a variable - the fact that it is a variable is likely to surprise people. It seems to be setting up for a problem like: if then delete = 'yes' ... if delete = yes then delete foo If I understand your example correctly, delete foo concatenates the values of the variable delete (which is 'DELETE' if hasn't been initialized) and foo, inserting an implied blank between them, then passes the result to an environment set up previously with an ADDRESS statement? This is what I want to avoid using SIGNAL ON NOVALUE. To me, it is much clearer to be explicit, including the concatenation, e.g. "DELETE " || foo seems much clearer about exactly what is happening/expected, which are variables and which are (expected to be) constant etc. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: REXX vs other languages WAS: Rexx numeric digits and scientific notation question
On 22/04/2024 10:09 am, Seymour J Metz wrote: It shortens expressions, upper cases automatically and, IMHO, is more readable. Was that a reply about the usefulness of the default value being the variable name, uppercase? If so, I still don't see it. Can you give an example of how you would use it? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: REXX vs other languages WAS: Rexx numeric digits and scientific notation question
On 22/04/2024 1:19 am, Seymour J Metz wrote: 2. There is a decades-long argument about NOVALUE. I am in the camp that believes the default behavior to be too useful to give up; others believe that it is dangerous. Genuine curiosity: what is the use of the default behaviour? I have never been able to see anything particularly useful. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: REXX vs other languages WAS: Rexx numeric digits and scientific notation question
On 20/04/2024 12:41 am, Rony G. Flatscher wrote: It is an attempt to apply dynamic typing to reduce the need to write the explicit type. This service gets carried out by the compiler. In true dynamically typed languages you can reuse variables like "start" or "tmp" to refer to values of different types. It's very specifically not an attempt to apply dynamic typing. It's type inference, and can only be used when the compiler can determine the type. The resulting type is still static. There's no practical difference between var start = ZonedDateTime.now(); and ZonedDateTime start = ZonedDateTime.now(); except for the amount of code that needs to be written. A dynamic language when executing code needs nevertheless to check whether interaction with the dynamically typed values is correct or not. This is one of the reasons I prefer static languages. Run time errors can go unnoticed until they are discovered on a customer system. Compile time errors are almost certain to be discovered in the development environment. (Or for in-house programming, compile time errors are discovered in development during business hours, runtime errors might be discovered in production at 3am...) Any error that can be found at compile time instead of runtime is a win. No the Java syntax is all in all much more complex. You note that when teaching novices Java and compare that to teaching novices Rexx or ooRexx (I have experiences in both). There might be a distinction between the difficulty of writing a program that runs, and writing a program that is correct. Well in Rexx concatenation rules are quite intuitive, surprisingly there is a blank concatenation operator (a blank between concatenated strings), an abuttal concatenation operator (no blank between concatenated strings) and using the explicit concatenation operator || (no blank between concatenated strings). Novices have no problems to understand: a=1 say a "- hello!" -- concatenation with a blank ...: 1 - hello! say a"- hello!" -- abuttal, concatenation without a blank: 1- hello! say a || "- hello!" -- concatenation with || operator ...: 1- hello! I had no problem back in the day when I first learnt it, but having learnt other languages since I don't think this is intuitive. Particularly the added blank between strings (I agree it is surprising). There are other variations here, e.g. I'm not sure how abuttal works with 2 variables rather than a variable and a literal. Rexx defines the value of variables that have no explicit value assigned to them to be the name of the variable in uppercase, so it does not cause a problem. (It does cause problems for programmers who got trained to always make sure that a value is assigned to variables. This is the reason why ooRexx has an option to activate checking for the use of uninitialized variables in Rexx programs to appease those who are not accustomed to it or feel that it should not be allowed. ;) ) It doesn't cause a problem except when it does. One danger is a typo in a variable name. When I wrote some complex Rexx I did use SIGNAL ON NOVALUE, but I wasn't totally appeased! I'm not trying to be negative about Rexx. I still use it when appropriate. But I think z/OS people make a mistake ignoring Java. It's fast (e.g. 1GB/s for my CICS SMF processing), makes powerful features available (e.g. the Twilio example to send SMS messages from z/OS) and it's available on pretty much every z/OS system now. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: REXX vs other languages WAS: Rexx numeric digits and scientific notation question
On 20/04/2024 1:42 am, Jay Maynard wrote: Agreed Java is simply far too complex a language and ecosystem to hold in the mind. Python is as ubiquitous and much easier to deal with. Really? What do you NEED to learn for Java that you don't need to learn for other languages? public static void main(String[] args)? Loops, if statements, data types, string processing are things you need to learn for all languages. I/O is useful (Rexx EXECIO isn't exactly intuitive). There are lots of add-ons, but you learn them (or just reference the details) if you need them. You still need to install add-ons and learn extra functionality for Python. Packages providing additional functionality are a good thing generally. The Twilio Python sample looks very similar to the Java version: https://www.twilio.com/docs/messaging/quickstart/python#send-an-sms-using-twilio I ran the Java version on z/OS by uploading the Twilio jar-with-dependencies and running the sample under BPXBATCH. Is Python easier??? //BPXBATCH EXEC PGM=BPXBATCH,REGION=512M //STDPARM DD * sh /usr/lpp/java/J11.0_64/bin/java /home/andrewr/java/src/TwilioTest.java "Hello from z/OS" //STDENV DD * CLASSPATH=/home/andrewr/java/lib/twilio-10.1.3-jar-with-dependencies.jar TWILIO_ACCOUNT_SID=xx TWILIO_AUTH_TOKEN= //SYSOUT DD SYSOUT=* //STDOUT DD SYSOUT=* //STDERR DD SYSOUT=* -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: REXX vs other languages WAS: Rexx numeric digits and scientific notation question
On 18/04/2024 8:29 pm, Rony G. Flatscher wrote: The mileage of people here vary including the Java people themselves who have started to reduce the need of explicit declarations like the new "var" (imitating JavaScript) instead of strict types or foregoing the static main method such that one can at least code the main method without the explicit declarations. The motivation about these changes is to make Java easier, reduce typing needs and the like. The Java var keyword is more like C# than Javascript. The variable still has a strict type - you can only use var if the compiler can figure out the type from other information e.g. var start = ZonedDateTime.now(); start is a ZonedDateTime. You can't use it before it is defined, you can't assign anything other than a ZoneDateTime to it, you can't create a new variable called start in the same scope, whether or not it is a ZoneDateTime. You can't e.g compare it to a LocalDateTime without specifying a timezone for the LocalDateTime - that is one of those things that helps avoid errors. var just reduces redundant code, e.g. specifying the type twice in the same statement. Of course a static and statically typed languages with a compiler must define as much rules as possible, such that the compiler can check for them all. The more rules the more time consuming and the more difficult to learn a language. I think the syntax rules for Rexx are actually more complex than Java, because it is less likely to flag an error if you do something that's not actually what you want. E.g. string concatenation where variables are expected to be strings but maybe not, might not be initialized, sometimes you need vertical bars but not always etc. If you're used to the language you write it without thinking and avoid the traps, but the rules are there nonetheless. Java is relatively straightforward, and shares many rules with other languages - C++, C# etc. I'm not saying Java is perfect - it has its own traps (int vs Integer etc) but I find it a much easier language to work with. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Anyone exploiting ZEDC?
On 18/04/2024 12:04 am, Michael Oujesky wrote: Just a thought, but anyone processing internally compressed CICS or DB2 data on a non-z/OS platform (Windows/Unix) might see substantial CPU usage from RLE decompression. If the compression is lightweight, decompression should be too. I can't speak for any other product, but I did an experiment with the EasySMF Java API. Running a CICS report on my laptop I got: Processing CICS compressed data: 1.2 GB/s (size after decompression) Processing uncompressed data: 800 MB/s So processing the compressed data was actually about 50% faster. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: REXX vs other languages WAS: Rexx numeric digits and scientific notation question
On 18/04/2024 4:39 am, Rony G. Flatscher wrote: As you know already Rexx it would be easy for you to learn about what ooRexx adds to Rexx. ... Notabene: you write one ooRexx program that will be runnable without any changes on Windows, Linux and macOS. This means you develop it e.g. on Windows at home and execute it in a Linux s390x subsystem at work and vice versa. ;) ... A dynamic and dynamically typed language as ooRexx allows to forgo many of the declarations a static and statically typed language mandates, thereby simplifying coding quite considerably. I find Rexx difficult because explicit declarations and static typing (as well as tightly controlled scopes) actually make programming easier, in general. They show up bugs in the code and make it easier to write correct programs. The IDE is also an important factor. I already write programs on my Windows laptop and run them on z/OS using Java :-) -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Anyone exploiting ZEDC?
On 17/04/2024 12:09 pm, Michael Oujesky wrote: Yes and zEDC poorly compresses internally RLE compressed records. I was surprised how well zEDC compressed the already compressed records. Using my data: zEDC alone : 52000 tracks CICS compression + zEDC : 22000 tracks zEDC seems to be biased towards speed rather than compression ratio, so maybe the RLE compression helps by packing more repeated bytes into whatever compression window zEDC uses? Plus CSRCESRV uses GP engine cycles That's true - CPU is probably more expensive than storage, so this could be just an interesting side-track. On the other hand, I think zEDC has to decompress and recompress the data for SMF dump etc. so CICS compression might save some overhead for SMF housekeeping type operations, reducing the amount of data going through zEDC? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: REXX vs other languages WAS: Rexx numeric digits and scientific notation question
On 16/04/2024 3:08 am, Jon Perryman wrote:
From a language standpoint, REXX is just another language but it's
real strength is it's environment integration. Instead of the caller
maintaining libraries, the environment automatically integrates with
REXX. For instance, REXX in the TSO environment, gives you access to
TSO commands (address TSO) and z/OS programs (address linkmvs). Start
ISPF and address ISPEXEC is available. ISPF option 2 gives you address
ISREDIT. SYSCALLS ON gives you address syscalls.
Rexx has better integration with z/OS, but Java has better integration
with the rest of the world.
I just wrote a piece about sending SMS text messages from z/OS:
https://www.blackhillsoftware.com/news/2024/04/15/text-message-alerts-from-the-z-os-smf-real-time-interface/
Spoiler: it's 2 Java statements using the Twilio API.
Twilio.init(ACCOUNT_SID, AUTH_TOKEN); Message message = Message.creator(
new com.twilio.type.PhoneNumber("+14159352345"), // to new
com.twilio.type.PhoneNumber("+14158141829"), // from args[0]) .create();
Twilio provide a library that can be used to send text messages from
z/OS. Amazon provide a library that can be used to work with AWS
services from z/OS. It's very common for cloud providers to provide Java
libraries for working with their services. Most of them will work on
z/OS and open up those features to the mainframe.
Java is also a much more powerful language. I used to write a lot of
Rexx, but I hate to go back because it is so much easier to get things
done in Java.
Rexx is good for small tasks where the overhead of starting the JVM is
significant, or where there isn't functionality in Java. Otherwise, Java
is my choice.
--
Andrew Rowley
Black Hill Software
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Anyone exploiting ZEDC?
On 17/04/2024 7:52 am, Michael Oujesky wrote: My current recommendation is to remove internal compression from CICS CMF (110/112) and DB2 (100-102) records, Zedc compress the SMF logstreams, then compress the logstream offloads via SMS zEDC compression. Have you compared compressed size of zEDC with/without the CICS compression? I have limited samples, but it looks like zEDC might compress data better if it has been already been compressed by CICS. Of course, there is the CPU time to consider. I think I/O can also be much faster for zEDC compressed data. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Rexx numeric digits and scientific notation question
On 19/03/2024 12:40 pm, David Crayford wrote: We utilize both languages, selecting the most suitable for each task at hand. Our primary application runs on Java using the Spring Boot framework. We orchestrate records originating from various z/OS data sources, transforming them into JSON or other pluggable formats, and dispatch them to analytics platforms, AI engines, or serve Prometheus metrics. We've developed numerous Java record mapping classes generated by our tooling, which is crafted in Python using Jinja2 templates and YAML schemas. Opting for Java in this context wouldn't be pragmatic. I've done JSON generation for all the SMF records supported by EasySMF in Java, so I guess it can be done either way. It certainly worked for me. With AI technology already a reality, Python stands as the prevailing programming language of the moment. While much buzz surrounds the new Telum chip in the z16, the question remains: How do we leverage its potential? For this, we require Python libraries—either TensorFlow or PyTorch—running on s390x architecture (for now). My impresson has been that Python is central to AI, but I'm curious about more general use cases. What you are using it for doesn't really tell me why you are using it. Is it a good choice for e.g. summarizing a few hundred million CICS SMF records? I guess AI implies processing large quantities of data so maybe it is. On the other hand, processing power tends to be less restricted on other platforms... -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Rexx numeric digits and scientific notation question
On 19/03/2024 11:05 am, David Crayford wrote: If you’re care so much about Java catching errors at compile time then why don’t you use Kotlin instead? NPE’s are one of the most common form of Java runtime errors that can be eliminated using Kotlins null safety. Java's not perfect, but it is powerful and it is pretty much universally available on z/OS. I'm writing functions for others to build on e.g.: https://static.blackhillsoftware.com/easysmf-rti/javadoc/com.blackhillsoftware.smf.realtime/com/blackhillsoftware/smf/realtime/package-summary.html So using/targeting the language that is most available makes sense. But my question was: Why Python? What are the reasons for using Python rather than e.g. Java? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Rexx numeric digits and scientific notation question
On 19/03/2024 10:01 am, Farley, Peter wrote: IMHO the learning curve for Java and all the libraries you have to know and understand to make it at all useful far outweighs its presumptive establishment in the z/OS ecosystem. You really don't have to learn a lot of libraries to use Java. Basic I/O and Java collections will get you a very long way. There are a lot of libraries, but they are there because they are (hopefully) easier than writing something yourself. You don't need to learn them until you actually need that function, and then you can make your choice whether to use an existing library or write your own. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Rexx numeric digits and scientific notation question
On 19/03/2024 9:53 am, Pew, Curtis G wrote: You get used to the “significant indentation” thing pretty quickly. It really makes sense: you should be indenting blocks anyway to make them easy to read, so why clutter things up with braces or keywords? Most scripting languages (including REXX) don’t require declaring variables. Since there’s no separate compile step you get the errors when they happen in any case, so declaring variables doesn’t buy you much. I probably would get used to significant indentation, although I do indent for readability and I'm not sure I want to be mixing readability and logic in the same construct. Explicitly declaring variables has a number of advantages: - it flags errors due to e.g. misspellings - it flags errors when you intend to create a new variable but reuse an existing one - it checks data types, e.g. for SMF data if you have a ZonedDateTime you can't assign/compare a LocalDateTime without specifying the time zone. So the class designers can protect you form a whole heap of common errors, and flag them before the program starts. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Rexx numeric digits and scientific notation question
On 16/03/2024 11:17 am, David Crayford wrote: IBM and ISVs are working on Python APIs for products right now. And they will be better than the REXX versions. Why Python? I know it's the latest hot language, but what advantages does it have over e.g. Java (well established on z/OS)? I have looked into it a few times, but get as far as "significant indentation" and "no variable declarations" and decide it's not something I want to use. Is it a sign of weakness to ask the compiler to help me avoid bugs? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: SSH tunneling for unattended process.
On 9/01/2024 6:26 am, Rick Troth wrote: It's been a minute, but I used SSH to carry PPP traffic back in the day. The client side PPPD ran 'ssh' as a child process, arranging stdin and stdout, as if it was a dial-up modem. I think this is difficult on z/OS because by default stdin and stdout use ASCII<->EBCDIC translation, so anything sending binary data is broken. I think you can override that, but not conveniently. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: ISPF to Panvalet interface
On 21/12/2023 3:29 pm, Brian Westerman wrote: compuware was acquired by BMC and they have decided that they want over $20K (close to 30) per year for just file-aid. Is that expensive? It doesn't necessarily seem expensive, compared to non mainframe software costs for the size of company that would run a mainframe. Especially when you consider limited market size etc. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Test site for certificate revocation?
On 20/09/2023 8:37 am, Charles Mills wrote: Does anyone know of a server URL that will present a revoked certificate (for my testing purposes)? Can you create a certificate for your own test site with Lets Encrypt, then revoke it? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Switching between SMT-1 and SMT-2
On 5/09/2023 9:19 pm, Scott Chapman wrote: It's more complicated than that. Although I would agree that if an LPAR has only a single zIIP, likely SMT would be a good idea. But B is not true for intervals that people usually consider when looking at utilization levels because at the level of dispatch intervals, it's much more likely there's at least some intervals where the zIIPs are 100% busy. It really depends on arrival patterns and how much you care about very short transactions that may be running on the zIIPs. I had a look at some RMF data, and I think that zIIP velocity (from zIIP using and zIIP delay in the SMF 72 records) might be a good indicator. It hopefully gives an indication of the real delay much better than the overall utilization. I saw that the velocities on my system were quite low i.e. lots of delay even with very low utilization. I suspect (guess) that this is due to short running Java jobs where the JVM is running multitple threads during JVM initialization. This would probably benefit from SMT2. On the other hand if your zIIP work is single threads arriving randomly you may not see delay until higher utilization. Does DDF run multiple threads for a single unit of work? I wrote a SMF real time interface sample to read type 72 records and issue a message or even a command to change the SMT setting based on theSMF 72 zIIP velocity across all service classes. You can see the program here: https://github.com/BlackHillSoftware/easysmf-samples/tree/main/easysmf-rti/rti-smt-switch My theory is if the zIIP velocity is low, the work would benefit from SMT2 because there are lots of threads waiting. If the velocity is high, SMT might be hurting due to multiple threads on a processor when there is no work waiting. Obviously what is considered "high" and "low" will depend on your environment. Martin's question about a WLM reset also applies. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Ray Mullins on Assembler demand.
On 7/09/2023 1:38 am, Seymour J Metz wrote: With that definition the learning curve is very different. You don't need to learn all the classes to use Java. It obviously helps if you know they are there, but if you know the commonly used classes and know how to find information when you hit a "how do I..." question you will be fine. Inheritance and interfaces help immensely, because many classes have common interfaces so you only really need to learn them if you need their particular specialization. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Switching between SMT-1 and SMT-2
On 6/09/2023 7:52 pm, Martin Packer wrote: I really hope you’re not advising customers to run the zIIP pool at 100%. Key functions such as Db2 DBM1 Deferred Write and Prefetch, as well as Db2 Log Writes, depend on very good access to zIIP. I'm not advising customers, I'm just saying that workloads like Java in particular can and will easily drive zIIP to 100%. I would have expected that WLM would be able to manage dispatching priorities to avoid impacting DB2. That might require e.g. the Java work running in a discretionary service class. Is that not the case? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Ray Mullins on Assembler demand.
On 6/09/2023 12:22 pm, Seymour J Metz wrote: What is "Java"? If it's just the language, it's not that big. But if it's all of the classes as well, ... I think it would have to include the libraries/classes that are included in a standard distribution. I'm sure there is a specification for what must be included in each Java version. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Is SMP/E needed for installs?
On 6/09/2023 9:39 am, Steve Thompson wrote: But, doesn't that take a long time and burn many cycles to get it done? I'm only asking to verify what I have been told about z/OSMF in the past few weeks. Good question. SMP/E itself doesn't have a reputation for being lightweight, and the z/OSMF install would avoid most of the SMP/E steps. Hopefully, a simple restore of a basic SMP/E environment in z/OSMF wouldn't be too intensive. Particularly compared to someone muddling through trying to understand SMP/E, re-running steps etc. For a product install, the CPU time etc. is probably not significant compared to everyday workload. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Switching between SMT-1 and SMT-2
On 5/09/2023 10:30 pm, Martin Packer wrote: And “100% busy” is the wrong criterion – if you want your system to be performant. Isn't that what WLM is for? 25 years ago I did a WLM conversion on a single CPU system. A single CPU is probably the most difficult to tune. By necessity, it ran at 100% for hours at a time. With appropriate WLM settings it performed fine. You do need a mix of work so you have work that can afford to wait, but it's not like the system needs time to to rest. There's no major difference between idle time and time running lower priority work. Is the situation worse now? There were a lot of things we had to do that went against the conventional wisdom for WLM at the time - the conventional wisdon did NOT work well at 100% or on a single CPU. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Switching between SMT-1 and SMT-2
On 5/09/2023 10:13 pm, Peter Relson wrote: I don't think that that is a correct conclusion. It is far more complicated than thinking that this is simply about one LPAR. It depends what your logical vs physical configuration is. It depends what resources your other LPARs need. And a lot more. I understand that this is designed to manage the dispatching between multiple LPARs. However, if the result is that experts advise to "only use SMT when [adding real zIIPs] was not (or no longer) feasible" and "not enable SMT until you have a defined reason to and where it's then proven to be beneficial", and customers are commonly seeing workloads degraded by enabling SMT, I'm not sure that the objectives are being achieved. Considering the interactions wth other LPARs, is there a significant difference between using the same dispatching pattern for SMT-2 as SMT-1 until all CPUs are busy, versus disabling SMT-2 completely? SMT should be all gravy. It's supposed to be extra capacity, and ideally undetectable until you hit the limits without SMT. And it is expected that zIIPs are not run "consistently 100% busy". That would not be an appropriate usage of zIIPs. Why not? Is WLM unable to manage dispatching priorities for zIIP? I know a lot of zIIP work is short transactions like DDF, but there's also Java, which is great for CPU intensive work. Java has functions that will check how many CPUs the system has, and parallelize work onto N-1 threads. It's a bit antisocial on z/OS I know, but that's what WLM is for. That would definitely drive zIIPs to 100%, potentially for extended periods. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Is SMP/E needed for installs?
On 6/09/2023 7:34 am, Phil Smith III wrote: To recap, the goal here is just to make our SMP/E installation as easy as possible for the increasing number of customers for whom it seems to be a mystery. Yes, that's worrisome; no, I can't fix it. What I can do is try to keep them from blowing their fingers off as much as possible by making it as easy as possible. IBM's answer to this would be to recommend installing using a portable software instance through z/OSMF. That will deliver the whole SMP/E environment pre-built, and the customer doesn't need to do any SMP/E for installation. They can still install maintenance using SMP/E, and you can be reasonably confident that the SMP/E environment is configured correctly so they can install maintenance. There are some downsides (new stuff to learn etc.) but given your goal and the fact that IBM is pushing people to z/OSMF installation anyway, it's worth considering. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Switching between SMT-1 and SMT-2
On 31/08/2023 10:50 pm, Scott Chapman wrote: But z/OS "densely packs" the cores, meaning that if a work unit is running on a zIIP core and another zIIP eligible work unit comes in it will run on the second thread on the already busy zIIP core instead of being dispatched to an available but unused zIIP core. As I understand it, this was done because PR/SM dispatches cores, not threads, to the LPARs and this dense packing makes that easier. Very interesting, I can see how that would cause problems. It seems like an attempted optimization that actually causes problems in the real world i.e. anyone not already running zIIPs at 100%. The solution seems simple - IBM could change the dispatching to work exactly the same for SMT-1 and SMT-2, until a work unit would have to wait. Then instead of waiting it gets dispatched as a second thread. Maybe you lose a few % at the top end, but it becomes something the customer can turn on and use without analysis. The current situation sounds like SMT-2 should only be used if you a) have a single zIIP or b) are running your zIIPs consistently 100% busy and for b) you need to turn it off when the workload reduces? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: [EXTERNAL] Re: Switching between SMT-1 and SMT-2
On 30/08/2023 10:22 pm, Horne, Jim wrote: Peter, In theory, I agree with you. In practice, I know of at least one shop with one system where overall throughput was worse with SMT=2 than with SMT=1 - and none of their systems got the benefits IBM predicted. I'm not saying it's common but it has happened. Do you know why, or the nature of the workload? It seems like it should be rare - until there are more processes than processors, everything should run at 100%. If you have more processes than processors, SMT means you have work being dispatched that otherwise would be waiting. Work that might be negatively impacted would be work that occupied close to 100% of a processor, and forced other work to wait. SMT would allow the other work to run improving overall throughput but reducing the speed of the first job. That's the theory anyway - what does it look like in practice? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: JAVA IDE
On 24/08/2023 3:11 am, Kirk Wolf wrote: To answer you question, for a real Java IDE, Java programmers generally believe that IntelliJ is the best and that's hard to argue with.I've used Eclipse for a really long time. If you are doing z/OS Java development, it's generally best to develop on your workstation and just deploy compiled jars to z/OS. I use Eclipse, but it's mainly inertia preventing me from trying something else. VS Code looks interesting because of the focus on supporting other z/OS languages, and the ability to edit datasets and files on the mainframe. I have been playing with Java single file source code programs under Java 11 on z/OS, and that's a nice facility. It makes Java more like e.g. Rexx, in that you don't have to run a compile step. Does anyone know whether it's possible to set a classpath and have auto-complete and syntax checking using VS Code to edit a Java file on z/OS? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: XCFAS and TRUSTED
On 21/08/2023 9:28 am, Lennie Dymoke-Bradshaw wrote: Secondly, when IBM states that a task should be given the attribute of Trusted, then I take it to mean that IBM is saying that the task can be trusted that this attribute cannot be the source of an exposure for that task. I think when IBM says a task should be given trusted, it's a stronger statement than that. I take it to mean that the task should never be denied access by the security system, and any denial of access risks the stability or operation of the system. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: XCFAS and TRUSTED
On 20/08/2023 8:53 pm, Mike Cairns wrote: I worked at one site many years ago where the local specialist had actually tested across multiple IPL's the necessity for each and every one of these tasks to actually have the TRUSTED attribute and the conclusion was that many of these did not actually need to be TRUSTED and could manage perfectly fine using normal RACF access to resources granted via permissions to profiles. I worked at a site which did a similar exercise. The risk is: 1) If the doc says it should be trusted, IBM are free to add functions that require access to other resources without documentating them. It's possible that IBM don't even consider what access would normally be required for an address space they specify as TRUSTED, or test it without TRUSTED. 2) There may be functions that are invoked only in unusual circumstances, so you only find out that access is missing when you are already dealing with a problem. Not worth the risk, in my view (our security group disagreed!) -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Looking for A Solution
On 21/08/2023 1:14 am, Steve Beaver wrote: On the zSeries I have created a .CSV. The issue is I need a solution to send this .CSV to a list of email Users that are very low skilled to their email accounts. And this Job needs to be hands-off so it can be scheduled. You probably want a solution that sends mail direct to your corporate emil system rather than setting something up to handle mail on z/OS. Connecting using TLS is probably also desirable or required. I would do it using Java. I wrote up an example a long time ago here: https://www.blackhillsoftware.com/news/2017/06/21/sending-email-from-zos-using-java/ There are a few changes I would make based on what I know now, but the basics are there. There will be plenty of examples on the internet showing how to add attachments to an email from Java. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: USS Features
On 16/08/2023 6:17 am, Jon Perryman wrote: This is absurd. Not all disk is cheap (e.g. GDPS). Not all data is valuable. While a person may be expensive, not everything they do is of value to the business and worth the hidden expenses. It's not cents per GB PC cheap, but it's not 1990s expensive either. If you have a meeting with a couple of people, schedule a change, get it approved, implement it... you could easily spend $1000 of people's time, not to mention add weeks to a project. How much disk space would that buy you, if you can avoid that cost and delay? The IBM RDP systems charge $10/month for 5GB of disk space, which seems expensive. But still, $1000 of meetings avoided will pay for 40GB of space for a year. What is the (ballpark) cost per GB on a real customer system? How much space is it worth to reduce the time spent managing it? You can't be serious about being a storage admin. Every situation and company are different. Questions must be asked. How do you not understand adding 100GB to a filesystem has an impact on GDPS, HSM, backups, recovery and much more. If you believe, everything is created equal, 100GB has the same impact on a 10GB or 10TB filesystem. A file system may contain millions of Unix files but its 1 MVS dataset. Recovery of a filesystem is risky at the best of times but add 100GB increases the risks and may impact the nightly archival time (1 Unix file change causes HSM to backup the entire filesystem). If as you say, data is valuable, then the UNIX backup would be used. I could go on but I expect this should be obvious. Whole filesystem backups are not very useful - really just a DR tool. What do you do if a user wants file(s) restored? Restore all their files from a point in time? Or restore the filesystem, mount it somewhere and manually copy files? The whole filesystem backup takes us back to the problems with individual filesystems. You are going to back up a user's whole filesystem, including all the freespace from the largest file they ever deleted, because they logged on and updated .sh_history? Or worse - can the filesystem change indicator tell the difference between a data update and a metadata (e.g. last accessed date) change? I'm not saying everything is equal, I'm just saying that freespace is a lot cheaper than managing a lack of freespace. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: USS Features
On 15/08/2023 10:09 am, Jon Perryman wrote: This is z/OS with SYSPROGS, not Unix with sysadmins where programmers have full control to define reasonable. You keep asking the wrong question. Who (not what) determines reasonable. Right or wrong, it is their job, not yours. If you can't give up control of sysprog duties, then z/OS is not the OS for you. I started on MVS in 1991, so it's a bit late to tell me that. I've *been* the storage admin. Without evidence to the contrary, its a good idea to assume the fact that your colleagues are being paid means that they are doing work valuable to the business, and it's not the storage admin's job to veto it because it requires too much DASD space. Its the storage admin's job to make sure that other people are not prevented from doing their work due to a lack of disk space. It's one of those jobs where the better you do, the less people know you're there. Disk space is cheap. Data is valuable. People are expensive. Don't waste expensive people time managing empty space. Spend your time looking after the valuable data, and have enough free space that people don't need to stop what they are doing and set up a meeting with the storage admin group. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: USS Features
On 14/08/2023 3:30 pm, Jon Perryman wrote: > On Monday, August 7, 2023 at 04:33:24 PM PDT, Andrew Rowley wrote: It comes back to the question I asked earlier - how much space is it reasonable to use *to do your job* before you have to get the storage admin involved? Since you put it that way, I've got to say are you insane. The storage admin based his decision on things called facts. He is responsible for storage and his decision is final until his management tells him otherwise. I suggest you inform your company CEO that all decisions must go through you. That was a question, not a decision. We all use space on the system, whether it be personal JCL libraries, sort work space, job output etc. If your manager says "Can you investigate this CICS problem from yesterday" and you want to analyze the SMF data, at what point do you need to ask the storage admin for space for sort work, data files etc? 100MB? 1GB? 100GB? I recognize that there has to be a limit where the storage admin has to be involved, but what is a reasonable value? Should it be different for a unix temporary file than for a sort work dataset? Maybe I decide the best way to investigate this problem is to generate JSON from the data and load it into Splunk. Is the disk space available? Or should I just download the CICS SMF data to the PC and process it there? Spool space, sort work space, user datasets - that space all comes from pools and when you delete the data the space is returned for use by other users. For some reason we set up a system for unix files where free space isn't automatically returned, and cannot be used by other users. This makes things unnecessarily difficult on the mainframe. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: USS Features
On 8/08/2023 12:56 am, Jon Perryman wrote: It's absurd to allow everyone to do Proof Of Concept on z/OS. Are all POC vital to the business? Are POCs disruptive to the business? "me" mentality ignores the impact on everyone else. In this case, you're saying the storage admin is not impacted when clearly that's not the case. It comes back to the question I asked earlier - how much space is it reasonable to use *to do your job* before you have to get the storage admin involved? There can be good reasons to do it on z/OS, e.g. the source data is already on the system, data security on other platforms etc. There is some irony in the contradiction between "z/OS because of its I/O capabilities" and "100GB! Whoa! Be reasonable!" -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Automount (was USS Features)
On 8/08/2023 12:37 am, Jon Perryman wrote: Automount was created specifically to address some filesystem blemishes. There's a problem they needed to solved and they allowed people to continue without the use of automount. For those who choose automount, they decided that with all its faults, it solved more problems than it created. IBM didn't create automount. It was a standard unix thing before IBM did unix. IBM just came up with the idea of HSM migrating home directories as a use case. The primary problem with individual filesystems is that freespace doesn't get returned to the system. Deleted a file? The space still can't be used by someone else. If you accidentally fill up your filesystem, when you delete the file after all those "growing filesystem" messages: congratulations, you own the empty space. The secondary problem is that migrating filesystems makes file and directory level management impractical. # du --sh /home/* # find /home -size 2G Don't even think about it, unless you like HSM recalls. File level backup also gets complicated when filesystems are migrated. Pretty much all the problems that automounted individual filesystems are supposed to solve are actually a result of having individual filesystems. They don't have to be solved on other platforms because they didn't create them in the first place (or there is a better solution e.g. quotas). -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Definition of mainframe? Was: Ars Technica
On 1/08/2023 12:16 am, Rick Troth wrote: But, again, an automount per user does not necessarily mean a filesystem per user. Agree... but I was specifically talking about a filesystem per user as a bad thing. This seems to have become a common thing on z/OS. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: USS Features
On 31/07/2023 10:59 am, Steve Smith wrote: Whatever. We use automount, and the "space" wasted is way too trivial to worry about. And HSM can magically free up home filesystem zfs files that aren't used any more. If it's trivial, you're probably not using actually using it. A low end laptop has 250GB available. How much space should a z/OS user be able to use (to do their job) before they have to make a special request to the storage management group? 10GB? 100GB? Some of my testing runs to (temporarily) 100GB+ for input and output files. I run it on the PC because the space isn't available on the mainframe, but It would be nice to be able to run it on z/OS. If you get a few users with usage spikes to 100GB the space might not be so trivial. gil answered that one... if you really have a good reason to go poking around in users' business. HSM recalls are the big problem with that. And authorized_keys is the sort of question where auditors might require you to be poking around in users' business. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Definition of mainframe? Was: Ars Technica
On 31/07/2023 9:36 am, Paul Gilmartin wrote: It mimics the MVS tradition of overallocating datasets. Aren't modern filesystems virtual and dynamically extensible? They do dynamicaly expand. It's not growing that's the problem though, it's shrinking - releasing space so that it can be used by another user. Hmmm... Use RACF to enumerate hone directories in OMVS segments. The "hard" then is just the performance cost of so many automounts. One of the claimed advantages of automount was that user filesystems could be migrated by HSM. Enumerating the users is fine if the filesystems are all instantly available. If you have a few thousand that need to be recalled from HSM... Does HSM release unused space from the filesystem when it is migrated and recalled, or do you need enough DASD to allocate all the empty space in all the filesystems? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Definition of mainframe? Was: Ars Technica
On 30/07/2023 2:28 am, Jon Perryman wrote: ASK YOURSELF: Name the z/OS Unix feature that sort of fixes the fundamental design flaw with Unix filesystems just described? I suspect most people won't think about each user having a unique filesystem using automount to make their filesystem available. Typical Unix uses one file system with all users having directories in the /user directory. An automounted filesystem per user has always been a terrible idea. I think it was given as an example of how you could use automount and somehow morphed into a recommendation. (Other OSes can e.g. use automount to mount a remote user filesystem via NFS). Reasons it's a bad idea: 1) Freespace in the filesystem is not shared between users. This means that you need much more space than if there was one pool of freespace shared between all. 2) It makes simple questions like e.g. "Which users have a .ssh/authorized_keys file?" much harder to answer. A filesystem per user is basically equivalent to a SMS storage group and catalog per user. You get isolation between users, but at the expense of much more difficult management. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Preferred FTP Client for Windows
On 27/07/2023 8:32 am, Schmitt, Michael wrote: FileZilla for GUI. Windows built-in FTP for command line, but... There are NO (free) command line FTP clients for Windows that support passive mode AND work with the MVS file system. Zilch. Nada. NOT A SINGLE ONE. For command line FTP on Windows, the Windows Subsystem for Linux (WSL) provides Linux versions of the FTP client. Much better than the original Windows version. I also use IBM Explorer for z/OS for transferring things to OMVS. I'm not sure how easy it is to set it up as a file transfer solution to MVS datasets. It requires a component set up on z/OS, but the advantage is it uses https connections. https://www.ibm.com/docs/en/explorer-for-zos -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Need DFSORT control statements to extract data from smf15 with storclas blank
On 24/07/2023 11:17 pm, Sri h Kolusu wrote: Was that a typo ? Isn't it supposed to be SMF15SCN ? or you are using SMF14 record layout with SMF15 Interchangeably as they are both have the same layout? The SMF 14 and 15 layouts are the same, so there is one implementation for both. This aligns with the IBM SMF macros. There is a Smf15Record class, but it just extends the Smf14Record class without adding anything. It's basically so you can create a Smf15Record object when you have a type 15 record, but the field names are still the smf14... names. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Need DFSORT control statements to extract data from smf15 with storclas blank
On 22/07/2023 12:46 am, shivang sharma wrote:
Hi ,
I am not a DFSort expert but we have a requirement to extract datasets
which has storclas (SMF15SCN) = blank.
Not DFSORT, but this is what it looks like in Java using EasySMF:
import java.io.IOException;
import com.blackhillsoftware.smf.SmfRecordReader;
import com.blackhillsoftware.smf.smf15.Smf15Record;
public class NoStorClass
{
public static void main(String[] args) throws IOException
{
try (SmfRecordReader reader = SmfRecordReader
.fromName("//DD:INPUT")
.include(15))
{
reader.stream()
.map(record -> Smf15Record.from(record))
.filter(r15 -> r15.smsClassSections().isEmpty()
|| r15.smsClassSections().get(0).smf14scn().equals(""))
.forEach(r15 ->
{
System.out.format("%-24s %-8s %s%n",
r15.smfDateTime(),
r15.smf14jbn(),
r15.smfjfcb1().jfcbdsnm());
});
}
}
}
If you want only new datasets (I suspect you do) you can add another
filter step:
.filter(r15 -> r15.smfjfcb1().jfcnew())
--
Andrew Rowley
Black Hill Software
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Problem running GIMUNZIP job.
Maybe this problem? https://community.rocketsoftware.com/discussion/miniconda-conflicts-with-bpxbatch On 20/07/2023 8:10 pm, Gadi Ben-Avi wrote: Hi, I managed to find the problem. My user was running a different shell which was causing the problem. Once I reverted back to /bin/sh, everything started working. Gadi -Original Message- From: IBM Mainframe Discussion List On Behalf Of Colin Paice Sent: יום ה 20 יולי 2023 11:35 To: IBM-MAIN@LISTSERV.UA.EDU Subject: Re: Problem running GIMUNZIP job. I do not see //SYSOUT or //SYSPRINT in your JCL... try adding them. //SYSPRINT DD SYSOUT=* etc You might try adding ECHO 'BLAH' to the script so you know something will be displayed Colin On Thu, 20 Jul 2023 at 08:46, Gadi Ben-Avi wrote: Hi, We are in the process of installing z/OS v2.5 from z/OSMF. One of the first jobs uses GIMUNZIP to create the new z/OS libraries. When the job got to the SMPPTS library, we got these messages: GIM43501S ** THE CALL TO THE BPX1WRT SERVICE FAILED WHEN PROCESSING /tmp/izud-V110AVN-T1314651/unzip/smpe2023201100726173157/S0354.CB.S0354.CB.ST252475.SMPE.SMPPTS.pax.Z. THE RETURN CODE WAS '0077'X AND THE REASON CODE WAS 'E329'X. GIM47800S ** AN ERROR OCCURRED WHILE GIMUNZIP WAS PROCESSING ARCHIVE ZSMSP9.CB.ST252475.SMPE.SMPPTS. GIM20501IGIMUNZIP PROCESSING IS COMPLETE. THE HIGHEST RETURN CODE WAS 12. Before the GIMUNZIP step, the job allocates a new ZFS and is supposed to mount it at /tmp/izud-V110AVN-T1314651/unzip, but from what I can see, this is does not happen, and the files are restores to /tmp, which does not have enough space. The job includes some shell scripts that are supposed to mount the ZFS, but it doesn't look like they are running. The step looks like: //MOUNTEXEC PGM=BPXBATCH,COND=(0,LT) //STDPARM DD * SH ; mpdir=/tmp/izud-V110AVN-T1314651/unzip; dsn='V110AVN.SWDEPL.T1314651.ZFS'; if -e "$mpdir" ¤; then; rm -r $mpdir; fi; if ! -e "$mpdir" ¤; then; echo "Work directory $mpdir will be created."; umask 077 ; mkdir -p -m 700 "$mpdir"; fi; PATH=$PATH:/usr/sbin; echo "Format the file system $dsn."; zfsadm format -aggregate $dsn; echo "Mount $dsn on $mpdir."; mount -s nosetuid -t ZFS -f $dsn $mpdir; There is no output in STDOUT or STDERR. How can I see the output of the script to make sure it is actually running? I would ask IBM, but this is running on z/OS v2.3, which is not supported, and they won't help. Thanks Gadi -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Python 3.11 on z/OS - UTF-8 errors
On 13/07/2023 2:00 pm, kekronbekron wrote: Andrew - could you please explain what you mean by "checkout in git as UTF8". I mean setting the zos-working-tree-encoding for the file to UTF-8. I don't actually do this but it appears to be possible. Git encodings are quite confusing. I find it frustrating that it seems to require you to tag every file added on z/OS, even if you have defined the encoding in .gitattributes. I did briefly try Maven on z/OS but I think I had the same problem as you. I quickly decided I didn't need to do that and haven't investigated any further. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Python 3.11 on z/OS - UTF-8 errors
On 13/07/2023 1:59 pm, Seymour J Metz wrote: UTF-8 is just an encoding of Unicode; not a character set. All of ISO-8859-1 is part of Unicode. Yes, but the code is stored in git as UTF-8 and defined as UTF-8 when working on other platforms. You can specify the zos-working-tree-encoding is ISO-8859-1... it seems like this conversion could cause problems. Alternatively, zos-working-tree-encoding of UTF-8 presumably transfers the data without conversion, but it sounds like some programs have problems using it? Normally, I think I want IBM-1047 which has the same potential conversion issues to/from git... you just have to be careful editing on other platforms. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Python 3.11 on z/OS - UTF-8 errors
On 13/07/2023 10:01 am, David Crayford wrote: We specify UTF-8 in our Maven builds as most of the time we are building off host on machines with UTF8 locales. However, we tag our files ISO8859-1 on z/OS ... If we cared about the euro sign we could change it to ISO8859-15 which is still an 8-bit character set. It’s those pesky codes above 0x7F in UTF-8 that cause the issues. Euro was just an example, there are plenty of other UTF-8 characters. If you convert to an 8 bit character set, does it mean that any literals with codes above 0x7F are silently broken? Or does git fail to checkout? Either way, sourceEncoding=UTF8 seems like a good answer to why you might want to actually have the files encoded in UTF8. Anything else would seem to be courting unpredictable errors. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Python 3.11 on z/OS - UTF-8 errors
On 13/07/2023 9:02 am, David Crayford wrote: I'd like to understand your reasons for wanting to encode your Java source files in UTF-8. It's important to note that the default encoding on z/OS is IBM-1047 (EBCDIC). We typically use ISO8859-1 and have to specify the "-encoding iso8859-1" option when using the javac compiler. As mentioned earlier, tagging files as UTF-8 can lead to unexpected issues, which is why it's not commonly done. I commonly see UTF-8 specified for Java projects. I have wondered how this works if you try to compile on z/OS. The obvious but possibly wrong answer would be to checkout in git as UTF8 and tag the files. If you encode as ISO8859-1, what happens to e.g. literals with characters not in ISO8859-1? An obvious one would be the Euro character, but I'm sure there are more. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: SAS Replacement
On 29/06/2023 9:12 am, Neil O'Connor wrote: Have a look at Black Hill Software for SMF reports/graphs. See https://www.blackhillsoftware.com/ Thanks Neil! There is also a Java API as well as the Windows software. The Java API allows you to write custom reports and run them on the mainframe or other platforms. The Java API can also convert SMF data to JSON format, so you can use any of the JSON reporting tools for queries, charts etc. eg: https://github.com/BlackHillSoftware/easysmf-samples/tree/main/smf2json -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
SMF Real Time Interface
I have created a Java interface to the z/OS SMF Real Time Interface, and I am looking for people interested to try it out. As background, the SMF Real Time Interface works in SMF logstream mode and allows you to define SMF in memory resources to receive different combinations of SMF records. Access to the in memory resources is RACF protected, and does not require exits or authorized code. The Java interface makes it very easy to connect, read and disconnect from the SMF in memory resources. I have put some samples on Github to show how it can be used: https://github.com/BlackHillSoftware/easysmf-samples/tree/main/easysmf-rti There are samples to show: * Simple usage: connect, read, disconnect * Send SMF records in binary format to another system via HTTP * Send type 30 job end information in JSON format to another system via HTTP * Send SMS text messages from the mainframe for job failures using the Twilio Java API. These will hopefully give some ideas about how it might be used. Andrew Rowley -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: [EXTERNAL] Re: z/OSMF
On 27/06/2023 11:35 pm, Pommier, Rex wrote: If a company is using a 20 MSU system and is forced to buy a 200 MSU system because that's the smallest available, unless something drastic happens to hardware AND software pricing, the company isn't going to be looking for "what else can I put on this expensive and woefully underutilized machine", they're going to be looking at "where can I move my 20 MSUs worth of processing to get rid of this expensive machine". Again, the assumption is that the pricing was adjusted so it was not more expensive. 200 MSU was just to pick a number. It probably should be higher. What is the speed of a desktop PC these days? I am guessing my 4 year old PC is around 600-700 MSU equivalent? A Raspberry Pi about 40-60 MSU? And 4 or more cores makes a lot of problems go away. It would be good if z/OS at the low end had kept up with the performance and price improvements in the rest of the industry. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: z/OSMF
On 27/06/2023 5:40 pm, Gibney, Dave wrote: 200 MSU was much more than we needed. Maybe 20 or so at our peaks. And the z/OS and ISV charges at 200 levels would have been intolerable. That's why I keep qualifying it with ISVs would have to adjust their pricing. While the smallest system is 13 MSU they have no reason to do that, but if the smallest available system was 200 MSU I think they would be forced to adjust. If you are using 100% of a 20 MSU system, z/OS is just an expensive system that can't do much more than run the stuff still around from the 90s. If you are using 20 MSU of a 200 MSU system, it's an expensive system hugely underutilized, and maybe people start asking what work can be moved there, how to make better use of the data it contains etc? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: z/OSMF
On 27/06/2023 3:07 pm, Timothy Sipples wrote: IBM doesn't require anyone to order/configure less than 200 MSUs (PCIs) of general purpose processor capacity. If you want to order a configuration like that go for it! Of course you can choose to configure larger machines, but these smallest configurations make no sense. They only exist due to software pricing. They are too small to seriously consider for any modern work, including it seems advances in z/OS itself i.e. z/OSMF. I mean, yes, ISVs and IBM can write software to run on them but why should they be so limited? Why not take advantage of advances that have happened on other platforms? You can't buy e.g. Intel systems with performance limited to the same as 20 years ago. If IBM made the smallest orderable system comparable to e.g. small-mid Intel systems customers would actually have some capacity to move new work to z/OS. Obviously software pricing would have to be adjusted so customers could afford it. I guess that's why IBM don't do it - no-one can figure out what to do with the pricing increments between 13-200 MSU. here's the current minimum orderable machine configuration (latest model) for z/OS and VSEn: * IBM z16 A02 (or AGZ for rack mount) * Capacity Model A01 * Base CP capacity: 105 PCIs (13 MSUs) I don't have access to this type of system to test, but it would be very interesting to run my Java CICS SMF reporting on this system and compare to a Raspberry Pi. I would have a small wager on the Raspberry Pi. Add just 1 zIIP and you get ~1,900 PCIs of full-time zIIP capacity with 2 processor threads (SMT2). You can add as many zIIPs as you wish up to the physical capacity of the machine. 1 zIIP is obviously much better for Java, but if we were to compare it to e.g. my laptop, my money would be on the laptop. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: z/OSMF
On 27/06/2023 11:52 am, Tom Brennan wrote: I haven't used z/OSMF myself so bear with me, but if it runs as a java application, doesn't that allow the possibility of running some kind of agent on the mainframe, and moving the heavy stuff (whatever that is) off to a 5 year old laptop? Not really - Java is just a programming language. Like any language, you can run part of your application here and part there if you build a communications architecture, have the data in the right places etc. But you can't just divide the application without doing the design work. z/OSMF does actually run part on z/OS and part in your browser (Javascript I assume) but the parts that access z/OS services, read z/OS data etc. need to be on the mainframe. I doubt there is much that can be practically offloaded. (On the other hand, I think the mainframe side performance could be greatly improved. Having looked at the SMF data from z/OSMF startup, I suspect the startup time could be cut by 90%+ with a bit of tweaking...) -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: z/OSMF
On 26/06/2023 8:13 pm, Timothy Sipples wrote: That said it's typically a "really good idea" to configure machines running z/OS with at least one zIIP — and not just for z/OSMF but for myriad other reasons. I've said it before but I'll say it again - to avoid embarrassment alongside 5 year old laptops or perhaps even a Raspberry Pi, IBM needs to figure out how to bring the smallest z/OS systems up to a modern configuration - I would suggest minimum 4 processors and 200 MSU. With z/OSMF, IBM has become a victim of their own policies. Software licensing has really strangled the low end of the mainframe market. IBM and ISVs need to figure out how to reverse this i.e. modified pricing so customers can move to larger machines. At a time when a laptop has 10 physical/12 logical processors, "one zIIP" isn't much better. zIIPs and System Recovery Boost seem to be a kludge to work around the dire lack of processing capacity in these systems anyway. (zIIP does benefit me because I write Java, but I can still see it's a bad solution.) -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: SMTP and OAuth
On 14/06/2023 11:57 pm, Tony Thigpen wrote: Asking for a VSE shop so he can answer an auditing 'request': Are there any SMTP clients on z/OS that support OAuth? I haven't used or tested it for OAuth, but Java usually provides an easy way to implement functions used on other platforms on z/OS. I would expect the JavaMail/Jakarta Mail OAuth examples to work on z/OS. https://javaee.github.io/javamail/OAuth2 It should be usable stand alone, and it looks like it is included in WebSphere Liberty? https://www.ibm.com/docs/en/was-liberty/base?topic=features-javamail-16 It depends what you mean by a client I guess, whether Java support counts or if you are looking for a pre-built product. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: The new requirement for Certificates to communicate with IBM -- A Journey
On 13/06/2023 8:04 am, Tom Longfellow wrote: I am beginning to suspect some new evil is afoot in the land of Java -- complete with unhelpful cryptic error messages. How old is your Java installation, and how old are the certificates required? It's possible that e.g. Java hasn't been updated to include new root certificates. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: The new requirement for Certificates to communicate with IBM -- A Journey
On 12/06/2023 2:59 pm, Tom Longfellow wrote: I am worn out from all of these "learning" opportunities and want to get back to "doing" the job I am paid to do. IBM doesn't do its customers any favors with the way they handle certificates. Every other operating system installs default trusted certificates, and all this "just works" (mostly). IBM has decided they don't want to tell customers who to trust, but the reality is that this makes communication over the internet difficult or impractical. It's even more difficult because RACF is different to everything else out there, so it's hard to find examples. I accept that IBM has customers who need to do their own vetting of CAs. However, IBM could provide e.g. a separate optional FMID that installed a set of trusted certificates updated by PTF, the same as other operating systems. Customers could select whether or not the standard CA certificates were installed. That would make life much easier for customers who just want to use TLS for internet connections, and leave vetting CAs to their operating system vendors. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: The new requirement for Certificates to communicate with IBM -- A Journey
On 11/06/2023 12:17 am, Tom Longfellow wrote: Has anyone managed to accomplish this using the power of the mainframe without ruffling the feathers of the Windows/Browser world? I have full certificate management powers under RACF on the mainframe. I just do not have a usable Certificate to Import and Add to the SMPE KeyRing. I wonder whether certificates can be exported from the Java keystore on z/OS for import into RACF? -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Are there samples of COBOL methods or classes?
On 29/05/2023 8:40 am, Tom Ross wrote: ... The real worl is not object oriented...we do no instantiate a customer object and then invoke a 'send_a_bill' method on that object. We procedurally decide that it is time to send a cusotmer a bill and do it! I'm not sure I would agree... it probably depends on what sort of code you work with, but I think it would be extremely common to have a customer object and to have a bill object. Then e.g. you might have dozens of ways line items might be added to a bill, but the bill checks its status when you try to add an item and won't allow it if the bill has already been sent. A program always has a sequence of actions i.e. a procedure, but with well designed object classes you don't need to know the procedures beforehand. Re-usable tested mehods that could be re-used to build applications it just not really happening! It's happening a lot! -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Digitally signed product software packages from IBM
On 26/05/2023 4:28 am, Kurt J. Quackenbush wrote: Glad to hear it works great and "management will love it." If you find value in this capability I encourage you to reach out to your other software providers and request they also start signing their packages. I know one in particular is already working on it, but not sure about the many others. What about non-SMP/E delivered software? What would be nice to see is a function where e.g. APF and linklist libraries at least were required to be signed. I know there was a discussion some time back on the difficulties with load modules due to reblocking etc. However, we can also sign things on z/OS e.g. SMF data. So you could have a local signing key usable for functions like the binder and IEBCOPY, and under certain conditions e.g. - all input is signed - IEBCOPY etc. is APF authorized the reblocked module is signed with the local key, maintaining a chain of signatures that can be validated back to the original package. Other components (panels etc.) would be much easier to validate a signature. So it would be nice to be able to look at everything and see that it is either unchanged from a vendor, or something modified locally. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: COBOL to dynamic DD name
On 29/04/2023 11:48 am, Paul Gilmartin wrote:
Obviously a complete Java program needs very few comments.
Fair point... although I would contend that a Java program without
comments is easier to understand than DFSORT control statements! Here is
a version with comments (again, untested):
import java.io.*;
import java.util.*;
import com.ibm.jzos.*;
import com.ibm.jzos.fields.*;
import com.ibm.jzos.fields.daa.*;
public class App
{
public static void main(String[] args) throws IOException
{
// define JZOS fields for the key, count and ddname
BinaryUnsignedIntField keyField = new BinaryUnsignedIntField(0, 2);
BinaryUnsignedIntField countField = new
BinaryUnsignedIntField(4, 4);
StringField ddField = new StringField(8, 8);
// variable to save the current header record
// I'm not sure I understand the input data i.e. why this is
// necessary but I think this matches the logic from the File
Manager
// sample
byte[] nextHeader = null;
// map of ddname -> OutEntry containing RecordWriter and count
Map outEntries = new HashMap<>();
RecordReader input = null;
try
{
// open the input dataset
input = RecordReader.newReaderForDD("INPUT");
int bytesRead = 0;
// create a buffer for the input record
byte[] record = new byte[input.getLrecl()];
// read records
while ((bytesRead = input.read(record)) >= 0)
{
if (keyField.getInt(record) == 0) // if key is zero
{
// copy and save the header record to a new byte array
nextHeader = Arrays.copyOfRange(record, 0, bytesRead);
}
else
{
// get the ddname from the record
// (this probably doesn't match the File Manager
sample -
// might need to concatenate/substring etc)
String ddname = ddField.getString(record);
// find the output entry for this ddname
OutEntry out = outEntries.get(ddname);
if (out == null) // doesn't exist: this is the
first record for database
{
out = new OutEntry(ddname); // create/open a
new entry
outEntries.put(ddname, out); // save it in the map
out.writer.write(nextHeader, 0,
nextHeader.length); // write the saved header entry
}
out.count++; // increment record count for this entry
countField.putInt(out.count, record); // update
sequence in the record
out.writer.write(record, 0, bytesRead); // write
the record
}
}
}
finally // close all files
{
if (input != null)
input.close();
for (OutEntry entry : outEntries.values()) // each output
writer
{
entry.writer.close();
}
}
}
/**
* A class to keep a RecordWriter and count for each ddname
* best practice would be to encapsulate with getters/setters
* but we're trying to keep things simple for the sample
*/
private static class OutEntry
{
OutEntry(String ddname) throws IOException
{
// open a RecordWriter for the supplied ddname
writer = RecordWriter.newWriterForDD(ddname);
}
int count = 0;
RecordWriter writer;
}
}
--
Andrew Rowley
Black Hill Software
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: COBOL to dynamic DD name
On 29/04/2023 6:29 am, Schmitt, Michael wrote:
I have an input file that contains thousands of records. They are in groups:
header record, then a bunch of segments all for one database name, then another
header, records for another database. But the same database can appear more
than once in the input.
Have you considered Java?
This is untested, and I don't know File Manager so I'm not 100% sure of
the logic or even the field definitions, but it should give you the
idea. Run under JZOS Batch Launcher so JCL DDs are available.
A complete Java program:
import java.io.*;
import java.util.*;
import com.ibm.jzos.*;
import com.ibm.jzos.fields.*;
import com.ibm.jzos.fields.daa.*;
public class App
{
public static void main(String[] args) throws IOException
{
BinaryUnsignedIntField keyField = new BinaryUnsignedIntField(0, 2);
BinaryUnsignedIntField countField = new
BinaryUnsignedIntField(4, 4);
StringField ddField = new StringField(8, 8);
byte[] nextHeader = null;
Map outEntries = new HashMap<>();
RecordReader in = null;
try
{
in = RecordReader.newReaderForDD("INPUT");
int bytesRead = 0;
byte[] record = new byte[in.getLrecl()];
while ((bytesRead = in.read(record)) >= 0)
{
if (keyField.getInt(record) == 0)
{
nextHeader = Arrays.copyOfRange(record, 0, bytesRead);
}
else
{
String ddname = ddField.getString(record);
OutEntry out = outEntries.get(ddname);
if (out == null)
{
out = new OutEntry(ddname);
outEntries.put(ddname, out);
out.writer.write(nextHeader);
}
out.count++;
countField.putInt(out.count, record);
out.writer.write(record, 0, bytesRead);
}
}
}
finally
{
if (in != null)
in.close();
for (OutEntry entry : outEntries.values())
{
entry.writer.close();
}
}
}
private static class OutEntry
{
OutEntry(String ddname) throws IOException
{
writer = RecordWriter.newWriterForDD(ddname);
}
int count = 0;
RecordWriter writer;
}
}
--
Andrew Rowley
Black Hill Software
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: zOSMF and zOWE for non-mainframers
On 1/03/2023 9:01 am, Andrew Rowley wrote: The Achilles heel for Java on z/OS seems to be dataset I/O. It's adequate, but feels like it should be faster. (I haven't actually done direct comparisons with other languages.) Following up an old post of my own here, because I did actually compare the JZOS RecordReader to other languages (C and Assembler). It turns out I was wrong, and the Java I/O using RecordReader seems to be close to the speed of Assembler and C. Certainly not slow as I thought. So Java is fast all round. Incidentally, zEDC compressed datasets give a big boost in I/O speed for reading SMF data in my testing, e.g. 2-5x faster. This was tested on the on the Dallas RDP system under VM - I don't know whether VM affects I/O performance. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Just PDSE
On 8/04/2023 2:23 pm, Brian Westerman wrote: Just so you know, it's not SYSPLEX that you need, it's GRS and you have have that with just a couple FICON ports. Are you sure about that? Regular PDS needs GRS to safely share between multiple systems, if that was all that was required no-one would have additional issues with PDSE. Just because it hasn't happened doesn't mean you are not vulnerable to data corruption with the right set of circumstances. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Unzip on z/OS ?
On 31/03/2023 8:05 am, Schmitt, Michael wrote: What if the requirement is to compress or decompress an area of memory, by calling something from a COBOL CICS transaction running on z/OS, in a way that it could be compatibly decompress or compress by a program running on another platform? For example: CICS > compress > send across some communication path > transaction on Linux > program > decompress > same data as started with The actual algorithm isn't critical, as long as it is compatible on both ends. The compressed data isn't in a file, nor are jobs being run. This is transaction processing. This sounds like a description of HTTP compression. I don't know the details inside CICS, but it seems likely that if "some communication path" can be HTTP this might be as simple as turning on compression in the HTTP server and/or requesting it in the HTTP client. I would expect that CICS Liberty would be able to do this. zOSMF Liberty certainly does. Liberty will use zEDC if available (zEDC doesn't work with Java 11 yet, only Java 8). -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Stop the ragging on COBOL please [was: RE: ASM call by value]
On 28/03/2023 9:34 pm, Seymour J Metz wrote:
I once found myself defending the common idiom
for (;;) {
foo;
}
as a perfectly clear DO FOREVER.
I'm not sure that it is completely clear, it depends on knowledge if
whether the empty statement evaluates as true or false - or just a guess
that do forever is more likely than do never...
Personally, I prefer something like:
while (TRUE)
{
}
--
Andrew Rowley
Black Hill Software
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: watching JIT generated machine instructions on z/OS
On 23/03/2023 11:39 am, David Crayford wrote: FYI, the OpenJ9 JIT can implement JIT intrinsics to generate native decimal instructions when running on z/OS. IBM have open source OpenJ9 and OMR so you can snoop around the code https://github.com/eclipse-openj9/openj9/blob/master/jcl/src/openj9.dataaccess/share/classes/com/ibm/dataaccess/DecimalData.java More interesting is: https://github.com/eclipse-openj9/openj9/blob/master/jcl/src/java.base/share/classes/com/ibm/jit/BigDecimalExtension.java since that implies DFP is used for BigDecimal calculations. However, BigDecimal is immutable so by specification calculations create a new object for every intermediate value. Hopefully the compiler can optimize away most of them, if you require maximum performance. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: watching JIT generated machine instructions on z/OS
On 23/03/2023 3:42 am, René Jansen wrote: Does anybody know how to ask the J9 (Java 8) on z/OS how to show me what it does when the JIT decides native code would be best? I think it is difficult to pin down the native code because the JIT is not a one-time, fixed output compiler. Java can monitor the execution of code and come back and perform further optimization on hot areas of code. The docs even talk about "optimistic optimization" where certain assumptions are made and the code can later be de-optimized if they turn out to be false. So trying to look at specific machine code seems like a problem. What types does the code use? I don't see a built in decimal type other than BigDecimal, so the opportunity for Java to use DFP instructions seems limited - but I don't really know. My inclination would be to try to get some performance figures for the Java code using e.g. JMH (which provides warmups etc. to trigger JIT optimizations) and try to build the equivalent the other language. Benchmarks have their pitfalls, but it's probably still better than trying to compare the native code. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Question for our international friends (mostly)
On 18/03/2023 11:54 pm, Bob Bridges wrote: Now that you mention this, I guess I'm being inconsistent when I say "see eye see ess" and "eye em ess", but privately chuckle when ignorami say "are ay see eff" instead of "rack-eff". RACF always amused me, because "rack off" is a (Australian?) way to somewhat rudely tell someone to go away. So, "I tried to log on, but RACF told me to rack off." It's an accidentally great name for a security product. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: not using SMPe
On 17/03/2023 9:20 am, Jeremy Nicoll wrote: But doesn't the vendor already have to understand whether or not separate changes they've made to their product affect each other? In theory yes, in practice you rely a lot on the compiler to verify calls between modules etc. It's much easier to do in source code than with object modules, and in high level languages the compiler does the checking much more accurately than any other process could. Shipping exactly what you built rather than individual pieces of it is much easier and less error prone. Obviously larger thing like an operating systems have components that are updated individually, but for a moderately sized self contained application full replacement makes a lot of sense. It's also usually less work for the customer - as long as the installation is well designed so customization doesn't have to be repeated. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: not using SMPe
On 16/03/2023 12:01 pm, Warren Brown wrote: Andrew, I have been out for a while. Is there a replacement for SMPE ? Warren There have always been ways to deliver software without using SMP/E e.g. datasets dumped to tape. As people have noted, you can also deliver what is effectively a full replacement via SMP/E. Serverpac etc. could probably be considered a non-SMP/E install, even though what you are installing is a SMP/E environment. But now, yes, there is an official IBM supported non-SM/PE installation method using a z/OSMF portable software instance. A PSI can also deliver a SMP/E environment of course. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: not using SMPe
On 16/03/2023 9:33 am, Ed Jaffe wrote: Not true. It simply means if you change the (non-fixed) interface, you must re-issue all affected components. OK yes... I did say that you could change the interface with corresponding pre and co-reqs. And if you just want to change one of the dependent components, you then need to go back and install the other fix, and end up changing all the other dependencies, plus anything they pull in etc... Not everyone loves chasing pre-req chains in SMP/E. Plus the vendor needs to track it all somehow and build those chains. That said, we have been having great luck lately using ++JARUPD instead of ++JAR for many of our Java-based components and people seem to like it! OMG WTF I hadn't discovered ++JARUPD. That sounds like a terrible idea. Java is a bit of a special case because it is JIT compiled, so you do get compiler optimizations from inlining etc. But I don't see the point of partial updates to jar files. You throw away all the checking that the javac compiler did when it compiled the code, and errors are more likely to slip through and be discovered by the customer rather than in the build environment. Even more so now with JPMS, which is complex enough just building a jar file. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: not using SMPe
On 15/03/2023 8:33 am, Jay Maynard wrote: That's because SMP/E and its power are only truly present in the z/OS and predecessors world. Everyone else thinks of applying maintenance as a matter of replacing the entire product, instead of individual fixes that are automatically maintained and managed. The problem with individual fixes rather than full replacement is that it means the interface for every individually replaceable component is fixed. The relationships between all components need to be tracked, and any interface change needs corresponding pre-reqs or co-reqs for everything that uses it. That had to be done anyway when everything was coded in assembler. With high level languages it becomes much easier to do at compilation time. The compiler will find the relationships and make sure all dependent changes happen. Separate modules also limit compiler optimizations, because calls between routines are one of the common opportunities for optimization (inlining etc). It's still possible to ship individual fixes if necessary, but that is done in source code using git etc., still shipped as a full replacement. You end up with a branch or tag in git that represents the exact code that a customer is running. There is much more overhead and opportunity for error shipping individual components than full replacement. With the bandwidth etc. available now full replacement makes much more sense. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: zOSMF
On 7/03/2023 2:43 pm, Mike Schwab wrote: a zIIP gives adequate performance, otherwise takes 90 minutes to start up, on z/OS 2.5.. Not disagreeing with other criticisms of zOSMF, but IBM and ISVs need to get together and find an affordable way to get customers onto systems with a modern level of performance. Modern performance probably means minimum of 4 CPs and 300 MSU. That would probably put them somewhere alongside low end desktop PCs. On my development z/OS system, which runs Java about the same speed as my laptop, zOSMF starts in about 1 minute. My software is Java so I love zIIPs, but in reality they are a band-aid over the real problem - that software pricing has held back performance to the point that small systems are disadvantaged compared to just about any other platform. zIIPs were probably a mistake - they allowed IBM to ignore the performance issues with small systems for far to long. It's ridiculous that mainframes require software designed for such small systems. We might enjoy the challenge, but it's just killing the platform. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Mainframe REXX (Re: Badmouthing Rexx and ooRexx - again (Re: zOSMF and zOWE for non-mainframers
On 3/03/2023 1:36 am, David Spiegel wrote: Every time I've watched mainframe Java since OS/90 V1, it has been painfully slow. Have you ever watched a WAS startup? ... It takes about 15 minutes on the fastest mainframes. I think that's a reflection on WAS, not Java. When I was playing with the software drag racing, I compared the times from SMF to get an idea of the Java startup etc. overhead. For a 5 second elapsed time of 5 seconds timed inside the program, SMF showed: C++ : 5.02 seconds elapsed, 4.95 seconds CPU time Java: 6.16 seconds elapsed, 5.24 seconds CPU time. So the overhead to start Java looked to be around 1 second elapsed, and 1/4 second of CPU time. After the startup, Java was as fast as C++ in this test. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: zOSMF and zOWE for non-mainframers
On 28/02/2023 11:28 pm, Seymour J Metz wrote: What about Java? How well does JIT work? Java and the JIT compiler is very good. It always depends on the benchmark. For some stuff I think it's close to C. JIT compiles to machine code, and there are optimizations available in a JIT compiler that are not available to a regular compiler so there's no reason it wouldn't be fast. If your CPs are not full speed the zIIP gives an extra speed bonus. There is a cost to start the JVM. On my system that seems to be roughly 1 second each time you run a Java program. This is where Rexx etc gain an advantage for small programs. The Achilles heel for Java on z/OS seems to be dataset I/O. It's adequate, but feels like it should be faster. (I haven't actually done direct comparisons with other languages.) -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: zOSMF and zOWE for non-mainframers
On 25/02/2023 8:03 am, Bob Bridges wrote: Oh, I was going to mention that surely allocating datasets, either in batch or TSO, has got to seem like one of the dumbest and most incomprehensible things we do on the mainframe, to a foreigner. Allocating datasets shouldn't be that hard to grasp. "Dataset" is actually a pretty common term and generally implies a collection of data with more structure than just a file. If people have worked with databases, presumably they have encountered the concept of defining the schema. A dataset is simply a collections of records (similar to a database table) where each record has a fixed length or limited variable length and the total collection has a size limit. Any structure within the records is up to the application. This is not an unusual construct. What is unusual is that z/OS uses such a structured format for almost everything, not that it exists. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: zOSMF and zOWE for non-mainframers
On 25/02/2023 4:23 am, Farley, Peter wrote: The other issue is CPU usage in today's frequently-CPU-constrained-due-to-cost mainframe shops. This is a major issue on z/OS today. My laptop has 10 cores/12 logical processors. I'm developing on the laptop for z/OS. I constantly have to be aware that my code needs to run in an environment with much less capacity. With CPU power in such abundance everywhere else, it makes perfect sense to spend CPU time to save developer time. A computing platform where CPU capacity is so constrained becomes hard to justify. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: determine job that created dataset?
If you are not into SAS, here is Java to do a similar thing using
EasySMF Java API:
public class DatasetAccess
{
public static void main(String[] args) throws IOException
{
List targetDatasets = Arrays.asList(
new String[]{
"SYS1.PARMLIB",
"SYS1.PROCLIB"
});
// SmfRecordReader.fromName(...) accepts a filename, a DD name
in the
// format //DD:DDNAME or MVS dataset name in the form
//'DATASET.NAME'
try (SmfRecordReader reader = SmfRecordReader.fromName(args[0])
.include(14)
.include(15)
.include(17)
.include(18)
.include(61)
.include(62)
.include(64)
.include(65)
)
{
reader.stream()
.map(record -> new AccessInfo(record))
.filter(accessInfo ->
targetDatasets.contains(accessInfo.dataset))
.limit(1000)
.sorted(Comparator
.comparing(AccessInfo::getDataset)
.thenComparing(AccessInfo::getSmfDateTime))
.forEachOrdered(accessInfo ->
System.out.format("%-44s %-24s %-10s %-4s %-8s%n",
accessInfo.getDataset(),
accessInfo.getSmfDateTime(),
accessInfo.getEvent(),
accessInfo.getSystem(),
accessInfo.getJobname()
));
}
}
private static class AccessInfo
{
public AccessInfo(SmfRecord record)
{
smfDateTime = record.smfDateTime();
system = record.system();
switch (record.recordType())
{
case 14:
case 15:
{
Smf14Record r14 = Smf14Record.from(record);
event = record.recordType() == 14 ? "Read" : "Update";
dataset = r14.smfjfcb1().jfcbdsnm();
jobname = r14.smf14jbn();
break;
}
case 17:
{
Smf17Record r17 = Smf17Record.from(record);
event = "Scratch";
dataset = r17.smf17dsn();
jobname = r17.smf17jbn();
break;
}
case 18:
{
Smf18Record r18 = Smf18Record.from(record);
event = "Rename";
dataset = r18.smf18ods();
jobname = r18.smf18jbn();
break;
}
case 61:
{
Smf61Record r61 = Smf61Record.from(record);
event = "ICF define";
dataset = r61.smf61enm();
jobname = r61.smf61jnm();
break;
}
case 62:
{
Smf62Record r62 = Smf62Record.from(record);
event = "VSAM open";
dataset = r62.smf62dnm();
jobname = r62.smf62jbn();
break;
}
case 64:
{
Smf64Record r64 = Smf64Record.from(record);
event = "VSAM status";
dataset = r64.smf64dnm();
jobname = r64.smf64jbn();
break;
}
case 65:
{
Smf65Record r65 = Smf65Record.from(record);
event = "ICF delete";
dataset = r65.smf65enm();
jobname = r65.smf65jnm();
break;
}
default:
{
event = Integer.toString(record.recordType());
dataset = "Unknown";
jobname = "Unknown";
break;
}
}
}
public String getEvent() { return event; }
public String getSystem() { return system; }
public String getDataset() { return dataset; }
public String getJobname() { return jobname; }
public LocalDateTime getSmfDateTime() { return smfDateTime; }
private String system;
private String event;
private String dataset;
private String jobname;
private LocalDateTime smfDateTime;
}
}
--
Andrew Rowley
Black Hill Software
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Are JNI required to be re-entrant and/or re-usable?
On 14/02/2023 1:53 am, Steve Austin wrote: Yes you are correct, and the assembler program would need to be added to the program class. I had had hoped to find documentation to confirm what I've seen and maybe find a way around it, but no luck so far. I don't have exact documentation to hand, but I would be surprised if it was not required to be reentrant. It's not something I have specifically considered - it it possible to write non-reentrant C code? (As distinct from thread safe.) I think your JNI program probably has to do the work of loading the assembler program itself for each call. OR Is it possible to recreate the assembler program functionality in Java? That is definitely more maintainable than JNI + Assembler, and might be easier - depending on the function. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Are JNI required to be re-entrant and/or re-usable?
On 16/02/2023 6:40 am, Steve Smith wrote: If you're confused, it's probable that you didn't read all his words. The OP did say that the assembler module is *known* to be not reusable i.e. each call needs a new copy. Speculation about *maybe* it's not reusable and whether that was intended seems to be confused. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: How to get MetalC "INLINE" report
On 11/02/2023 7:42 am, Farley, Peter wrote: I'd consider this to be a bug in the compiler. The listing pmap should show all source statements, at the place where they are executed. If they are inlined twice, they should be listed twice. That sounds like it would ultimately be confusing. An inlined piece of code might be inlined in hundreds of places. In general, my expectation of optimization is that it happens behind the scenes and doesn't change visible behavior. A huge growth in the compiler listing is probably not what people want. Also, once a piece of code is inlined, other optimizations are opened to the compiler. - The compiler might know some input parameters, so parts of the code are redundant and can be removed, loops can be unrolled etc. - Statements can be reordered - parts of the inlined code might be moved outside loops in the calling code - The compiler might even swap the order of inner/outer loops, so that part of the calling code ends up inside a loop in the inlined section of code These would be very difficult to reflect in a compiler listing (and are the reasons debugging optimized code is problematic). -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Transmitting SMF records
On 26/01/2023 10:40 am, Kirk Wolf wrote: ISTR that distributed SAS has an option to read binary byte stream files with BDW+RDWs, which is what you would get if the program were to produce a merged RECFM=U DCB. It does, but I can't see any reason to implement that other than to workaround IBM's broken FTP. It's easier to read data in record format with RDWs. IBM FTP will do the right thing if you specify the SITE RDW option - I assume that the SAS RECFM U support was added before that was available. EasySMF will automatically recognize RECFM=U, RECFM=V or TSO TRANSMIT format data, optionally in a gzip or zip file, and process it correctly. It will also recognize SMF data with no RDW and tell you what is wrong. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Transmitting SMF records
On 26/01/2023 10:20 am, Paul Gilmartin wrote: It's possibly catering to the novice user rather than the journeyman. Unfortunately I don't think that's the case. When transferring variable length binary records, the default options give you unusable (I would call it corrupted) data. You need to delve into the details of record formats and RDWs to understand why. If using the data requires deblocking variable records from RECFM U blocks it's not for the novice user. More likely, the people writing FTP didn't properly understand the requirements for variable length records. You need the RDW. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Transmitting SMF records
On 26/01/2023 7:24 am, Andrew N Wilt wrote: The GDKUTIL utility does allow an Upload of something with RECFM=U. This was done at the behest of a customer that was using ftp to put SMF data out for processing by the SAS tool. Apparently, they are able to decipher the RDW records resulting from uploading a RECFM=VBS as a RECFM=U. I would be slightly surprised if SAS can't read VB records properly formatted with the RDW. RECFM U was a workaround for the problem where FTP stripped out the RDW. So GDKUTIL seems to be reproducing the brokenness of FTP, then providing the same workaround as a feature. It would be preferable if it retained the RDW in VB records by default. GDKUTIL seems new enough that the default could be changed without too much impact. I doubt there is anyone using variable length binary transfers currently who needs the data without the RDW. (However there might be people uploading data who haven't discovered the resulting data is unusable.) I have been working with SMF data myself for many years, and this problem with FTP causes a LOT of customer confusion and frustration. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Transmitting SMF records
On 20/01/2023 7:07 am, Glenn Wilcock wrote: Late to the party on this discussion... DFSMS recently shipped a new utility named GDKUTIL. It converts z/OS data sets and UNIX files to cloud objects and visa-versa. Reading the documentation, the following note looks like a problem: = /When specifying a z/OS Data Set for LOCAL or LOCNAME for a DOWNLOAD command without the CONVERT keyword, the Record Format should not be Variable. (RECFM=V or RECFM=VB) This is because in binary mode, there are no indicators in the data where a record ends, so the data is placed up to the maximum logical record length. If a Data Set with variable records was uploaded to a Cloud Object in binary mode, the indicators of where a record begins and ends are lost. Therefore, a download operation of that same data to a Variable record data set cannot tell when to start a new record in the data set./ = A useful function should be able to round trip the data to and from cloud storage and end up with exactly what you started with. If there is not enough information to recreate the records on z/OS with the correct length, there is probably not enough information to usefully process the data on another platform. FTP made the same mistake, assuming that the record length information was not part of the data. For variable length records it is critical. RECFM U presumably preserves the block & record length information, but makes it unnecessarily complex to decipher. From the documentation I can't tell whether uploading RECFM U data results in the same data you started with. At a minimum there must be complications to end up with the real/correct DCB attributes. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Transmitting SMF records
On 16/12/2022 11:33 pm, Scott Chapman wrote: We have customers (S)FTP(S)ing us data every day directly referencing the output of IFASMFDP with DCB=RECFM=U with no problem. The OP specified being able to reverse the process, so my understanding was it needed to transfer back to a dataset on z/OS. This seems to be surprisingly difficult - IBM doesn't seem to have considered round trip capability when they wrote the FTP functions. (Although I haven't tried it with a RECFM U transfer.) -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Transmitting SMF records
On 16/12/2022 1:43 am, Gary Weinhold wrote: As an alternative, it wouldn't seem very difficult to create a utility to read the FTPed data (received as a sequential file with an arbitrary record length) and reformat the data to write a VB(S) file. It is not difficult. The biggest problem is that by default FTP strips the RDW, which makes variable length binary data unusable. The solution to that problem is to use the SITE RDW option for the download. I wrote a Java program to read/write z/OS data in GZIP format to get around the various problems. You can: - gzip VB(S) into a FB dataset on z/OS, download the data and unzip on a PC - download VB(S) data using BINARY and SITE RDW options, gzip the data on the PC, upload to a FB dataset then uncompress to a VB(S) dataset. The program is available here: https://github.com/BlackHillSoftware/z-java/tree/master/java/compress -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Capture 3270 screen as text (was: TNZ 3270 Emulator: ... )
On 21/11/2022 3:52 am, Paul Gilmartin wrote: That's likely not to work because lines might be separated and indented not by CRLF and leading spaces but by cursor addressing commands which are not meaningful when pasted into a text document. By the time the screen is displayed the 3270 commands are likely replaced by spaces in a fixed width font so copying text shouldn't be a problem. Vista does this very well. I don't remember any issues using PCOM either. It's more complicated if you want to keep the colors, and I don't recall an emulator that does that. Pasting is more complicated in that there are areas you can't write to. That's more a design issue though - how you handle multiple lines, do you wrap etc. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: TNZ 3270 Emulator: Any Experiences?
On 19/11/2022 8:19 pm, Peter Sylvester wrote: It seems to me that this code currently poses a security risk. Certificate validation always say OK. (well it is said like this in the code.) Which terminal emulators actually validate the certificate? I would like to think all of them do, but the number of TLS tutorials that start by creating your own CA suggests that certificate validation might not be common. I don't think Vista does it :-( I assume that TLS is set up before any TN3270 specific stuff happens so you can probably test it by connecting to e.g. wrong.host.badssl.com port 443. A certificate error would show that the certificate is being validated. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Dcollect to Splunk
On 10/11/2022 8:27 am, James Cradesh wrote: Is it possible and any vendors supporting it? Would like to get rid of SAS. It's not a released feature yet, but EasySMF:JE can convert Dcollect records to JSON for reporting with e.g. Splunk. Contact me direct if you would like to try it out. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: ftp to/from USS with Windows
On 26/10/2022 8:47 am, Steve Thompson wrote: So this is what I want to do: [already set to the correct location in windows] get '/u/abc/some.java.file.jar' some.java.file.jar Try without the quotes? I think the quotes might indicate a dataset rather than HFS file. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
Re: Enterprise Cobol 6.3.0, XML PARSE and zIIP usage.
On 30/09/2022 7:24 pm, Massimo Biancucci wrote: After increasing the size of XML up to 150MB then I saw a little of zIIP usage. Anyway, it would be interesting to know how IBM manages these thresholds. Interesting question. It might be as simple as whether the program is processing XML when it is interrupted and redispatched. The likelihood of being interrupted in XML code probably corresponds to the proportion of time spent processing XML, so dispatching on the zIIP would probably give you roughly the right amount of zIIP time even if you have some leakage of XML/non-XML code between zIIP and CP. -- Andrew Rowley Black Hill Software -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
