Guys,
We have recently changed from MLC (monthly) licencing of our software to SCRT (Sub-Capacity Reporting), which is supplemented by our z13s sending TSAD data automatically to IBM.
We of course collect SMF data and automatically dump via the user exit to full dumps to GDG datasets on disk, and now specifically just type 70 and 89 SCRT records to separate datasets in a separate GDG.
But, the question is - how best to collect the monthly SCRT SMF records such that we have a neat collection for each month's reporting period (2nd of previous month to the 1st of this).
We have no automation tools, and so anything must be via JES2 automatic commands, or some other easy-to-use mechanism.
How do YOU do it?
A secondary question is then how do you get the SCRT data (generated via the SCRT Java-based tool into a PDS) emailed to the relevant IBM address?
I would be grateful to hear of any ideas you people have, as I'm sure it has been a problem for most of us Sub-Capacity Licenced IBM Customers?!
Joined: 06 Jun 2008 Posts: 8697 Location: Dubuque, Iowa, USA
We use SMF logstreams and dump the 70 and 89 type records by month. We run this data through the SCRT program to generate the CSV (it takes care of the dates for us). We then use a batch job to email the CSV to IBM and FTP to an internal server for retention. The FTP uses our job scheduler to set the year / month for uniqueness and the email is done through SAS (which is easier). Without any such tools, you might want to download the report to a local server and run cron or Windows Scheduler to rename the download to include the year / month and move it into an archive directory.
We don't use the Java SCRT at this point. Yes, it will be required by October but we're waiting as long as possible.
Robert,
Thanks for the reply.
Yes, we have the SCRT program setup to generate the CSV, but need to get something in place to automate the emailing of the CSV file to IBM - saving the admin person (non-mainframe!) in the office from having to do it! :-)
Also looking to FTP to a SAMBA server, but might have to put that out to a different FTP server at this time, as SAMBA (our documentation server) needs some extra stuff configured in to allow that?
I guess my main question then is how to you split off the monthly SMF type 70 and 89 records?
Right now, the userexit kicks in on an SMF switch to dump to next GDG level, but it's how best to accumulate and stores the relevant month's worth that's proving to be a pain right now?
Joined: 06 Jun 2008 Posts: 8697 Location: Dubuque, Iowa, USA
I've used SAMBA on the mainframe; once the TCP/IP piece is set up right, it becomes very easy to move stuff between the mainframe and network.
You can email the CSV directly from the mainframe with appropriate SMTP code. There are a number of examples of sending attachments through SMTP on this forum.
We create a data set with the SMF 70 and 89 record types just for the monthly report and reuse the data set every month. With SMF logstream, we just use RELATIVEDATE(BYMONTH,1,2) to get the month's data. If you're still using data sets, read in the last 6 weeks and keep the 70, 89 records, then let the SCRTTOOL program get the correct starting and ending dates.
Joined: 28 Aug 2007 Posts: 1742 Location: Tirupur, India
Quote:
" If you're still using data sets, read in the last 6 weeks and keep the 70, 89 records, then let the SCRTTOOL program get the correct starting and ending dates."
I think this is a clever idea to feed the tool more data, so it eliminates the need to worry about the right starting point of the previous month.
At our shop we have scheduled jobs which collect SMF daily onto a GDG and on 2nd of the month, there is another job which reads all GDG and creates a monthly dataset and clears the GDGs.
Quote:
We don't use the Java SCRT at this point. Yes, it will be required by October but we're waiting as long as possible.
Robert, You could try the desktop version of the Java tool as an intermediate workaround until the Mainframe one is setup.
We don't have Java 7 on Mainframe, so limping along with the desktop version until we find some time for Java upgrade.
As for the automation of SCRT job itself, My SCRT job looks like below.
Joined: 06 Jun 2008 Posts: 8697 Location: Dubuque, Iowa, USA
Quote:
Robert, You could try the desktop version of the Java tool as an intermediate workaround until the Mainframe one is setup.
We don't have Java 7 on Mainframe, so limping along with the desktop version until we find some time for Java upgrade.
We've got Java 7.1 installed on the mainframe, but we've been using the non-Java SCRT for quite a while and we hear the Java version adds a lot of CPU time to the process -- especially since we don't have a specialty engine installed. We're talking about installing / testing it in September (and hope IBM offers another extension).
Robert, You could try the desktop version of the Java tool as an intermediate workaround until the Mainframe one is setup.
We don't have Java 7 on Mainframe, so limping along with the desktop version until we find some time for Java upgrade.
We've got Java 7.1 installed on the mainframe, but we've been using the non-Java SCRT for quite a while and we hear the Java version adds a lot of CPU time to the process -- especially since we don't have a specialty engine installed. We're talking about installing / testing it in September (and hope IBM offers another extension).
Robert,
As an aside, yes, I can confirm that the Java-based SCRT takes a lot of CPU time under the default WLM, taking the CPU up to 100% even at the quietest of times!
Vasanthz,
Yes, I will be looking at hgetting the email SMTP setup too, as this is long overdue. ;-)
javascript:emoticon('')
I like the idea of reading in more than you need, as I also noted that SCRT automatically seems to work out the dates.
But how do you automate your monthly collections? I am wary of just one automatic task to do this, as the machine might be down, or something could prevent it from working as designed. I have thought of using a control file in the monthly dump process to check if its already been done or not, but I'd be interested in how you manage it yourselves?
As I say, we have no real automation tool product installed.
Joined: 06 Jun 2008 Posts: 8697 Location: Dubuque, Iowa, USA
I assume you generate daily SMF data sets. Do you collect these into a weekly data set? If not, you might want to consider that -- create a GDG with the level count depending upon how long you want to keep your SMF data handy. With weekly SMF data sets (which could be on tape or DASD), then you just read in the (0) through (-5) generations to the SMF dump program, filter out the 70 and 89 record types, and run them into SCRT. You could also write a short program in the language of your choice to set the monthly start and end dates for the dump program and run 6 weeks of SMF data into the dump using the generated parameter. I think you could even create the parameter in Unix System Services and move it to z/OS -- I'll investigate more when I have a chance.
Even without automation, z/OS does have some tools to help make tasks run easier.
Thanks Robert.
Really appreciate your ideas here! If you can find anything more or have any questions, then by all means PM me if more applicable?
We currently dump our SMF datasets automatically when full/switched via the IEFU29 exit. As such, we don't have daily dumps but the above generates one a day on average, sometimes more, sometimes less.
My current idea is to amend our SMF dump job to append to an accumulated weekly SMF GDG dataset, then run a JES2 automated nightly command to submit a job which interrogates the dayofweek using REXX, returns that as condition code then conditionally executes or not to bump the GDG generation to +1 in readiness.
The following step (again conditional) would then take your suggestion to run SCRT against gens -5 through 0.
Joined: 06 Jun 2008 Posts: 8697 Location: Dubuque, Iowa, USA
That's how I see it working, and how we used to do it. As I said, we've gone to SMF logstreams which stores the SMF data for us and we use the logstream dump program to extract what we need (this utility allows extraction of data by week or month; there may be other options but those are the ones we use the most).
Thanks, I'll look at the LOGSTREAM option again at some point, but I'm guessing it must store horrendous amounts of data in the logstream datasets (couple, linear datasets?)
Joined: 06 Jun 2008 Posts: 8697 Location: Dubuque, Iowa, USA
When you set up the logstream, you tell it how much data to store online so you do have some control over it. We chose to keep plenty of SMF data available.
')
