Mainframe Path Start learning free
Applied7 min readLesson 3 of 3

Investigations and reporting what you find

Three investigations come up again and again: a data set that disappeared, a job that used too much CPU, and failed access to sensitive data. The method is the same each time, and so is the rule for the report: say what the records show, with times and IDs, and leave conclusions about intent to the people who own them.

1. The data set that disappeared

  1. Extract type 17 (scratch) and 18 (rename) for the window when it was last known to exist.
  2. Find the record for the data set name: it gives the job name, the time and the system.
  3. If the job name is a TSO user ID, the user deleted it interactively; otherwise look at that job's type 30 records and its JCL.
  4. If the data set's RACF profile audits successful access, type 80 records add the user and the access level.

2. The job that used too much CPU

  1. Extract type 30 subtype 5 (job end) for the period, and subtype 4 (step end) for the jobs of interest.
  2. Compare CPU time, not elapsed time; check zIIP time separately if the site uses zIIPs.
  3. Compare with the same job on previous days: a sudden increase usually follows a change in data volume or program.

3. Failed access to sensitive data

  1. Extract type 80 for the window and filter on the resource name.
  2. List each event: time, user ID, resource, access requested and access allowed.
  3. Look for patterns, such as repeated attempts by one user ID or attempts by many.
  4. Report to the security team; they decide whether it is a mistake or an attack.

Reporting

WriteAvoid
"Type 17 shows PROD.GL.MASTER scratched at 16:47:53 on MPS1 by job MPU0412.""MPU0412 deleted it on purpose."
"Three failed READ attempts by MPU2231, 09:14 to 15:02.""MPU2231 is trying to steal payroll data."
"GLPOST used 21 minutes 55 seconds of CPU, up from about 12 on previous days.""GLPOST is broken."
TRY IT YOURSELF

Which type 30 subtype is written at the end of a job?

Show a hint

Step end is 4.

Show the solution

Subtype 5.

Common mistakes

Comparing elapsed time when the question is CPU

A job can run long while using little CPU, waiting for I/O or other work.

Stating intent

Records show actions, not reasons. Leave conclusions about intent to security and management.

Leaving extracts lying around

SMF extracts contain sensitive information; protect and delete them.

What you will see at work

Key terms

Check your understanding.
Take this lesson's quiz and save your progress. Free.

Take the lesson quiz
← Inside an SMF recordBack to SMF in practice: answering questions with SMF