IBM z/Transaction Processing Facility

TPF : Library : TPF Newsletters

Products > Software > Transaction Systems > TPF > Library > TPF System Newsletters >

TPF HMC Operating System Messages Support (APAR PJ28622)
Roger VanDongen, IBM TPF Development

With the introduction of APAR PJ28622, a new type of console support has been added to TPF that allows the system to use the Hardware Management Console (HMC) as the prime CRAS. The HMC can serve as the system console for various operating systems such as z/VM, z/OS, and now TPF. When you select the Operating Systems Messages icon on the HMC desktop and associate it with a logical partition (LPAR) that is running a TPF system, the console window, a graphical user interface (GUI) type of console window, can be used as the prime CRAS for that particular TPF system. The HMC hardware is included as standard hardware that is shipped with IBM zSeries processors. The Support Element is included on the processor frame and has the HMC software installed on it. Workstations with the HMC software installed can be configured as part of the HMC network to control the mainframe, and also can serve as the prime CRAS.

TPF HMC operating system messages support was designed to allow you either of the following options:

IPLing the TPF system with the TPF HMC operating system messages console as the TPF system console immediately
IPLing on an existing 1052/3215 console connection and then using a ZACRS command to fall back the prime CRAS to the TPF HMC operating system messages console.

To configure your TPF system to allow it to write to the TPF HMC operating system messages console immediately from an IPL, an optional parameter, HMC, has been added to the CRASTB SIP macro. The options for setting this parameter are YES and NO, with NO being the default. The following shows an example of how you could code the CRASTB macro:

CRASTB PRCRS=(56,56,1F,56,56,56,56,56,56,56,56,56,56,56,56,56,5
6,56,56,56,56,56,56,56,56,56,56,56,56,56,56,56),
ROCRS=(010002,1052,B,
010002,1052,C,
010002,1052,D,
.
.
.
010002,1052,O),
ROCRS16=(010002,1052,P,
010002,1052,Q,
010002,1052,R,
.
.
.
010002,1052,6),
ALTPC=(1F,1F,56,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1
F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F,1F),
NCONSL=NO,
CTKC32LC=YES, GENERATE CTKC IN 32LC FORMAT
HMC=YES

In the previous example, the system programmer has indicated that the TPF system is to write to the TPF HMC operating system messages console as the prime CRAS immediately from an IPL (IPLB into CTIN into Restart and 1052 state and higher).

Note also that if you want to use the TPF HMC operating system messages console as the TPF prime CRAS, you must code the CRASTB macro for 1052 console support. The HMC only supports 3215 mode. If you are using 3270 consoles for the system prime CRAS, you cannot define the system to also use TPF HMC operating system messages support. If you coded the CRASTB macro with both NCONSL=YES and HMC=YES, an error message would result while running the system initialization procedure (SIP) to generate your TPF system.

Even if you decided to code your CRASTB macro with NCONSL=NO (1052 console support) and HMC=NO (do not use the TPF HMC operating system messages console window as the prime CRAS immediately on IPL), the TPF system would use your 1052 console as the prime CRAS when you IPLed your TPF system, and you could then fall back to the HMC at a later time by using the ZACRS command. The ZACRS command has been modified to accommodate manual fallbacks to and from the TPF HMC operating system messages console. However, the restriction that the system has to also be generated with 1052/3215 console support still applies. The following are some examples of how to fall back to and from the TPF HMC operating system messages console:

To fall back to the HMC from your 1052/3215 console, you could enter the following two commands:

zacrs fbk prc 010000 type-hmc cpuid-b opt-val
CSMP0097I 07.27.26 CPU-B SS-BSS SSU-HPN IS-01
ACRS0068I 07.27.26 HMC VALIDATED+
zacrs fbk prc 010000 type-hmc cpuid-b

The following messages would be displayed on the TPF HMC operating system messages console, and this console would now be the prime CRAS:

DSID0001I 07.27.26 THE SYSTEM ID IS DANBURY.CH0001 B
ON PROCESSOR FF02153F MODEL 2064 LOW CPU ADDRESS 0000+
CSMP0097I 07.27.26 CPU-B SS-BSS SSU-HPN IS-01
ACRS0066W 07.27.26 CONSOLE ALTERED, THE HMC IS NOW THE PRIME CRAS.+

To fall back the prime CRAS from the TPF HMC operating system messages console to the 1052/3215 console, you could enter the following two commands:

ZACRS FBK PRC 010000 TYPE-CON CPUID-B OPT-VAL
CSMP0097I 07.27.26 CPU-B SS-BSS SSU-HPN IS-01
ACRS0067I 07.27.26 PRIMARY 1052 VALIDATED+
ZACRS FBK PRC 010000 TYPE-CON CPUID-B

The following messages would be displayed on the 1052/3215 console, and this console would now be the prime CRAS:

DSID0001I 07.27.26 THE SYSTEM ID IS DANBURY.CH0001 B
ON PROCESSOR FF02153F MODEL 2064 LOW CPU ADDRESS 0000+
CSMP0097I 07.27.26 CPU-B SS-BSS SSU-HPN IS-01
ACRS0045W 07.27.26 CONSOLE ALTERED, 1052 SUB CHANNEL 056 IS NOW THE PRIME CRAS.+

(Note: The actual ACRS0045W message ouput does not wrap around on the console as it does in the example).

Additionally, a fallback can be done from the TPF HMC operating system messages console directly to an alternate CRAS defined in the CRAS status table, and from an alternate CRAS directly back to the TPF HMC operating system messages console. There is additional logic added to the ZACRS command processor that checks to make sure that the device type for the ZACRS validate command agrees with the device type of the ZACRS fallback command. The following examples show both the error condition and a good fallback scenario:

ZACRS FBK PRC 010000 TYPE-HMC CPUID-B OPT-VAL
CSMP0097I 07.27.26 CPU-B SS-BSS SSU-HPN IS-01
ACRS0068I 07.27.26 HMC VALIDATED
ZACRS REP PRC 010000 TYPE-CON CPUID-B
CSMP0097I 07.27.26 CPU-B SS-BSS SSU-HPN IS-01
ACRS0070E 07.27.26 REP CONSOLE TYPE DISAGREES WITH CONFIGURED TYPE.

ZACRS FBK PRC 010000 TYPE-HMC CPUID-B OPT-VAL
CSMP0097I 07.27.26 CPU-B SS-BSS SSU-HPN IS-01
ACRS0068I 07.27.26 HMC VALIDATED
ZACRS REP PRC 010000 TYPE-HMC CPUID-B
CSMP0097I 07.27.26 CPU-B SS-BSS SSU-HPN IS-01
ACRS0000I 07.27.26 COMPLETED

In the previous example, once message ACRS0000I has been generated, the prime CRAS is back on the TPF HMC operating system messages console. The ZDCRS command output has been modified also to indicate the correct device type for any system that is running with a TPF HMC operating system messages console, as shown in the following example:

ZDCRS
CSMP0097I 07.27.26 CPU-B SS-BSS SSU-HPN IS-01
DCRS0000I 07.27.26 CRAS STATUS TABLE
TERMINAL TERMINAL PRINTER
ADRS CPUID TYPE ADRS

RO 010002 B HMC
PRC 010000 B HMC
RO 010002 C HMC
PRC 010000 C HMC
RO 010002 D CON-1052
PRC 010000 D CON-1052

As the previous example shows, you can have a mixture of 1052/3215 consoles and TPF HMC operating system messages consoles in your loosely coupled complex.

It is possible to use TPF HMC operating system message console support on a VM test system. Effectively, VM simulates the TPF HMC operating system messages console from your VM virtual console. Basically, you are sharing the virtual console between VM and TPF the same way you can share the TPF 1052/3215 console with the VM virtual console. When running a TPF test system on VM, messages sent from TPF to the HMC look identical to messages sent from TPF to the 1052 console, and it can become confusing when testing fallback scenarios between the HMC and the 1052 console. You can adjust the colors of your VM system to help differentiate the output for the HMC from the output for the 1052 console by entering the following two CP commands:

'CP SCRE CPOUT GREEN NONE'
'CP SCRE VMOUT RED'

Using these definitions, messages to the HMC would be displayed in green and messages to the 1052 console would be displayed in red.

Additionally on VM, when you are entering commands from the virtual HMC, you must prefix each command with the following text:

#CP VINPUT VMSG

So, to enter a ZDSYS command from the HMC when the HMC is the prime CRAS, you would then enter the following

#CP VINPUT VMSG ZDSYS

If, by mistake, you just enter ZDSYS, the command will not be accepted because VM will attempt to enter the command to the TPF system via the 1052 console interface. Because the 1052 console is not mounted to the TPF system when the HMC is active (and vice versa), the ZDSYS command actually would not be read by the TPF system and actually would be left queued up by VM for the TPF system to read. If you enter a second command without the #CP VINPUT VMSG prefix, you may get a NOT ACCEPTED message in the VM status area, and once you fall back to the 1052 console, you will need to do a plain <ENTER> to clear up the NOT ACCEPTED hang from the 1052 input queue.

For the purpose of our testing in the TPF development lab, we used an IBM 2064 eServer zSeries 900 processor with two service elements attached, and we had additional remote workstations that ran the HMC software and were connected into the support element on that processor. The performance results from the TPF HMC operating system message console indicated that the HMC workstation will be able to keep up with existing console traffic provided that you keep a couple of issues in mind: Our effective throughput to the TPF HMC operating system messages console was about 16 I/Os per second. If all the messages were single-lined messages, this then equated to 16 lines of text per second to the TPF HMC operating system messages console.

Within our TPF test system environment, because of multi-lined messages coming out to the console, we would see an average of 3.2 lines of text per message block (our blocking factor), so we were seeing effective performance numbers of roughly 50 lines of text per second to the TPF HMC operating system messages console from our TPF system. We had compared this throughput to the actual throughput of some of the larger TPF customers. The throughput numbers that the customers had recorded on their system RO logs indicated that the throughput to the system console ranged as high as 114,500 lines of text in a 24-hour period. The blocking factor (the average number of lines of text per message sent) was between 3.77 and 4.1 lines of text per send. For a console log of 114,500 lines of text in a 24-hour period, this means that the system was averaging about 1.32 lines of text per second, or roughly one I/O per every 3 seconds. The highest hour of traffic reported was 6655 lines of text, or roughly 1.85 lines of text per second. The highest 2-minute burst of traffic in the 24-hour period was 538 lines of text, or roughly 4.5 lines of text per second. With a blocking factor of 3.77, the TPF HMC operating system messages console would be able to accept approximately 60 lines (16 x 3.77) of text per second, which is in line with the actual performance that we observed while writing continuous multi-lined message blocks to the TPF HMC operating system messages console.

Interestingly, when we ran multiple TPF LPARs and had them stream messages to the HMC console window, the performance of the TPF HMC operating system messages console output for each LPAR was inversely proportional to the number of LPARs actually writing to the TPF HMC operating system messages console. When two LPARs wrote, the performance of each was about 8 I/Os per second per LPAR; with four LPARs writing, it was 4 I/Os per second per LPAR; and with nine LPARs, it was approximately 1.8 I/Os per second per LPAR.

The chance for a backlog of messages can occur if several LPARs share an HMC and each send a burst of messages concurrently. Based on our observations, such a condition could occur with three or more LPARs sharing the HMC.

The introduction of TPF HMC operating system messages support gives you more flexibility for consoles to use as your system console. Additionally, this support gives added reliability to your TPF system by adding another layer of console fallback support, both for automatic and manual fallbacks. This support also gives you a wider range of connections to use when linking automation platforms, such as the TPF Operations Server, into your TPF system; and because the Hardware Management Console comes standard on any zSeries processor, it is not necessary to purchase additional hardware to take advantage of the console support with TPF.