Recovery Routine/Abend exit to release ENQs during abends

Siva NKK Kothamasu · Posted: Tue Feb 25, 2020 4:43 pm

Good Day,

We have a COBOL program that calls an HLASM program to take ENQ.

Once control is back into COBOL, it will call n number of other COBOL and/or ASM programs and releases the ENQ (by calling the same ASM pgm) before GOBACK.

This is an online program and the issue we have is, if there is an abend in any of the programs that the COBOL program calls, the ENQ is still held by the region causing transaction queue build-ups until the region is cycled.

Can this issue be solved using Recovery Routines (IEAARR) or having an ABEND EXIT is the only solution or are there any better alternatives.

Appreciate your assistance.

Apoorva · New User Joined: 28 Jan 2020 Posts: 49 Location: India

Yes you could have an ESTAE recovery routine to perform "DEQ" and other clean-ups. Your ASM program, that obtains the ENQ, could establish an ESATE once "ENQ" is successful, and could remove the ESTAE routine once "DEQ" is successful. So, if there are any abends, in between ENQ/DEQ, your ESTAE will get control which could perform the DEQ.

Siva NKK Kothamasu · Posted: Tue Feb 25, 2020 7:01 pm

Thank you Apoorva ji.

I will try it and let the forum know the outcome.

enrico-sorichetti · Posted: Tue Feb 25, 2020 7:03 pm

Apoorva · New User Joined: 28 Jan 2020 Posts: 49 Location: India

Siva NKK Kothamasu · Posted: Wed Feb 26, 2020 6:51 pm

steve-myers · Posted: Wed Feb 26, 2020 9:09 pm

A recovery environment is active until it is discarded.

There are two things confusing about recovery. I just mentioned the first thing. The other thing is there are two parts of a recovery environment.

The ESTAE exit routine
The exit routine receives control early in ABEND. Among other things it has to decide whether it's possible to recover smoothly, and do other things.
The recovery retry routine.

Now, to add to the confusion is a fact most beginners are ignorant about, but you have to know about this to understand what you can or cannot do in ABEND recovery. In OS/360 derived systems like z/OS programs running in task mode have to have a TCB, which everyone knows about, but less well know is a program must have a "request block" (RB). We see these RBs in ABEND dumps, but we generally do not understand their significance. As a program executes it can gather a number of these RBs. Two things must be present for a program to run.

There must be a TCB that is not blocked for some reason AND
The most recently acquired RB for the task must not be in wait. The status of the older RBs is immaterial

Now, when your program ABENDs, the system obtains an RB for ABEND. In order to run the ESTAE exit routine, ABEND creates yet another RB for the exit routine.

The environment you see in an ABEND dump has the RB for your program followed by the RB for ABEND and, finally, the RB created by the SNAP macro which is used to print the dump. In other words, SNAP, like your exit routine, is sort of a subroutine of ABEND. In an OS/360 ABEND dump this was clear. z/OS complicates this picture with a lot of extraneous c****.

Now let's go back to the ESTAE exit. After it does its thing it returns to ABEND. If the exit wants to continue running the program that failed it has to tell ABEND where the "retry" program is. You use the SETRP macro for this or you can alter a control block called the SDWA which ABEND obtains before it enters your exit routine yourself which is what SETRP does. If there is a retry program ABEND alters the request block that failed to go to the retry program and terminates the ABEND request block. At this point the system enters the "retry" routine using the RB that failed.

Now, if I understand what you want to do you use the ESTAE exit to DEQ the resource you want to release and return to ABEND and forget the "retry" routine. All this should work.

Apoorva · New User Joined: 28 Jan 2020 Posts: 49 Location: India

"Will the Recovery Routine be still active during the execution of Program C?"

Yes Siva recovery routine will be always active until you issue "ESTAE 0" which removes the most recently added recovery routine from the stack.

Suppose if you issue below two ESTAE followed by "ESTAE 0", "RCVY-RTN2" will be removed from the stack,

ESATE RCVY-RTN1
ESTAE RCVY-RTN2

Siva NKK Kothamasu · Posted: Thu Feb 27, 2020 12:47 pm

Siva NKK Kothamasu · Posted: Thu Feb 27, 2020 12:50 pm

Apoorva · New User Joined: 28 Jan 2020 Posts: 49 Location: India

You are Welcome Siva!

Siva NKK Kothamasu · Posted: Mon Mar 16, 2020 6:03 pm

I have coded the ESTAEX Routine to remove the ENQ and it is working perfect WHEN INVOKED BY OS for an ABEND.

Thank you for your support so far, Greatly appreciated.

However, at our shop, we have a product called HOGAN and if I activate my ESTAEX from any program running under Hogan, my ESTAEX was not being invoked by MVS.

However, even when under Hogan, my ESTAEX was being invoked for a S0C4 and the abend I am forcing using, ILBOABN0.

Apoorva · New User Joined: 28 Jan 2020 Posts: 49 Location: India

Siva NKK Kothamasu · Posted: Thu Mar 19, 2020 8:15 pm

Apoorva · New User Joined: 28 Jan 2020 Posts: 49 Location: India

Thanks for the update Siva! Looks like you already have a plan, and I am sure you will get it resolved.

[/quote]

I am new to ASM and this is the 1st time for me to write ESTAEX routines.

[/quote]

Good start Siva!