Batch job to issue commands at regular intervals..

mymf doubts · New User Joined: 12 Jul 2012 Posts: 29 Location: India

G'day All,

I am looking for a batch job that gets submitted and keeps running unless someone cancels/IPLs. The job has to issue SDSF commands at periodic intervals (like DA) or SMS commands like /D SMS,LIB(ALL). I am trying to gather the output of certain commands thru the day and use it for our healthcheck reports.

Thru /D SMS,LIB(ALL),DETAIL at regular intervals I am trying to see the consumption of VTS drives and present the availability thru the day.

Please help.

I am assuming the batch job will run REXX to do this...hence posting it here!

Nic Clouston · Posted: Wed Oct 03, 2012 6:11 pm

Are your system programmers aware of this attempt to destroy the useability of the mainframe? Have you asked if it is possible to get this sort of data from SMF data?

Robert Sample · Posted: Wed Oct 03, 2012 7:03 pm

The information you are wanting is already collected by the system in the SMF records. There is ABSOLUTELY no reason for you to reinvent the wheel, and there are many reasons for you NOT to proceed any further with your wrong-headed idea of a batch job.

Talk to your site support group about what is already available and use it.

Pedro · Posted: Wed Oct 03, 2012 9:16 pm

Nic Clouston · Posted: Wed Oct 03, 2012 9:22 pm

hogging an initiator all day - not the destruction of Planet Earth! OTT? Sure

Pedro · Posted: Thu Oct 04, 2012 1:36 am

dbzTHEdinosauer · Posted: Thu Oct 04, 2012 4:16 am

yes, much better to have a rexx script running all the time as a started task,
(since it is looking for resource intensive tasks, I wonder if it will report itself?)
instead of learning how to use SMF records.

Pedro · Posted: Thu Oct 04, 2012 5:39 am

Dick, I am not sure what parts of your post are sarcasm and which are not.

I suppose the wisdom (or not) of it depends on the definition of 'health checker'. I thought it was more of a real time system problem notifier thingy. In that case, it is not clear to me that examining SMF records is either timely or would consume less resources.

If 'health checker' is more of a search for historical data, the SMF is likely better.

dbzTHEdinosauer · Posted: Thu Oct 04, 2012 7:09 am

you are correct; it is sarcasm.

i just feel that rexx is not really a good choice for a real-time
system monitor.

not that rexx would be slow processing system commands,
i have build rexx monitors for db2 that dealt with a select set of packages.
also, it was a terminal issued cmd (rexx script)
that was 'controlled' at a tso session level.

but in order to accomplish what the TS wants,
requires accessing all jobs, does it not?

that amount of i/o would, in my view, be rather resource intensive.

The TS wants to monitor the Virtual Tape System, real-time.
so, what activities would be invoked real-time?
i.e. based on what is presented, what would the observer do
in the event something was 'wrong'?
as well as, what would be considered (wrong),
what would require real-time response - and what would the response be?

PeterHolland · Posted: Thu Oct 04, 2012 1:34 pm

Using System Automation/Netview, timers can be defined to issue the necessary commands.

mymf doubts · New User Joined: 12 Jul 2012 Posts: 29 Location: India

Sorry, for not giving the complete picture earlier!

Issue we are trying to address: We run a large datacenter for many clients, the SMF data of all those client machines get FTPed to our billing mainframe (to tapes). We have total 48 VTS drives on billing mainframe. At times, the FTP from client systems fail issuing 'NO UNIT AVAILABLE'!

Our plan: To try and capture the available VTS drives (/D SMS,LIB(ALL),DETAIL) at 30 min intervals and project a trend for VTS drive usage . Depending on the usage trends, try to spread the FTP process to times when there are more drives available. I have spoken to our performance team (SMF people), ignorance or incompetency or truth, they said no such information can be extracted from SMF logs! I tried PMR to IBM, they have no such reports!

Action so far: I have manually issued (/D SMS,LIB(ALL),DETAIL) and /D U,,ALLOC,B700 (B700 series - our VTS UCBs) at regular intervals and found many jobs that were eating up drives (not using UNIT=AFF)..some jobs were allocated up to 6 drives! We have corrected those jobs and seen drastic improvement in drive availability with this lil effort.

Now, I want to monitor the drive availability numbers every 30 mins. I do not have access to any automation tools, and my earlier post showed the best I could think...if any have better ideas without needing any monitoring tools....please help.

Nic Clouston · Posted: Thu Oct 04, 2012 4:40 pm

So is PeterHolland's suggestion a starter? Discuss with sysprogs or whoever supports netview (or equivalent).

vasanthz · Posted: Thu Oct 04, 2012 5:48 pm

Hello,

PeterHolland · Posted: Thu Oct 04, 2012 6:11 pm

vasanthz

I understand the TS wants ad hoc information, SMF is working with after the fact data. So Nic Clouston is right to tell the TS to talk with sysprogs and eventually SA/Netview people.

vasanthz · Posted: Thu Oct 04, 2012 7:39 pm

Hi Peter,
I guess you are right.

dick scherrer · Posted: Thu Oct 04, 2012 7:52 pm

Hello,

When severeal of my clients wanted to repeatedly capture some info like this, psocesses were scheduled to run periodically and write the output from that execution to the +1 of a gdg.

When the process to analyze the data was scheduled, the entire gdg was copied to another file and the original deleted. No worry about a tape drive always being available.

mymf doubts · New User Joined: 12 Jul 2012 Posts: 29 Location: India

Hi Dick,

The drive availability issue is not at the client machines, it is our billing mainframe that is facing drive availability (with 48 drives!). The billing mainframe receives (FTP) SMF records from many client machines throughout the day, we receive them onto tapes...they are v large to be put on DASD. At times, there are one too many FTPs (and batch jobs) running and the system runs out of tape drives.

I see many suggesting to try Netview, I am assuming it is part of every mainframe! Let me try finding someone with those skills here in our team!

dick scherrer · Posted: Fri Oct 05, 2012 9:29 pm

Hello,

As this is already "the way", there would probably be no re-implementation, but do consider that with 48 tape drives you are in this situation. When we created hundreds of new generations of gdgs on dasd daily, we did not have this problem.

Pete Wilson · Posted: Tue Nov 06, 2012 2:20 pm

Why not increase the number of virtual drives? I realise it may need an IODF change but it may help you.

In terms of reporting capability, check out the SMSVMA product supplied by IBM. It was originally developed in support of Tape Mount Management (TMM) where tape files were transferred to DASD by SMS. It can give drive usage profiles by hour as well as alot of other tape related information including jobname and program name etc. You basically feed SMF info into the first part to extract the required records, and then the subsequent part reads the extract to produce the report. I last used it about 15 years ago before there were VTS's so not 100% sure how it works for VTS but it was a very useful tool.

publib.boulder.ibm.com/infocenter/zos/v1r11/index.jsp?topic=/com.ibm.zos.r11.idak100/amsr11.htm