IMS in production
IMS is operated with slash commands to display and control transactions, programs and databases. Recovery depends on image copies and logs recorded by DBRC.
IMS commands
IMS commands start with a slash and are entered from an IMS master terminal, an automation tool or the z/OS console (with the right routing). Most can be abbreviated to three letters.
| Command | What it does |
|---|---|
| /DISPLAY TRANSACTION name | Shows a transaction's status, queue count and scheduling details |
| /DISPLAY PROGRAM name | Shows whether a program is started or stopped |
| /START TRANSACTION name | Allows a stopped transaction to be scheduled again |
| /STOP TRANSACTION name | Stops scheduling; messages queue up |
| /DBRECOVERY DATABASE name | Takes a database offline to online IMS (for example before recovery) |
| /START DATABASE name | Makes the database available again |
When an online program abends
If an MPP abends, IMS backs out the database updates and the input message for that unit of work. Depending on definitions, IMS may stop the transaction or program so the same message does not abend again and again. Messages then queue, visible in /DISPLAY TRANSACTION output. After the fix, operators restart with /START, and the queued messages are processed.
Backup and recovery
- Image copies are backups of databases, taken with IMS utilities on a schedule.
- Logs record every change; online logs are archived to system log datasets.
- DBRC records which image copies and logs exist in the RECON datasets, so a recovery can be built from the right inputs.
- Forward recovery restores the last image copy and reapplies logged changes up to the point of failure.
Reorganisation
As segments are inserted and deleted, IMS databases fragment and slow down. DBAs reorganise them on a schedule, unloading and reloading the data in hierarchical order. HALDB partitions can be reorganised one at a time, and some sites reorganise online.
Common mistakes
If a bad message caused the stop, restarting immediately can abend the program again and back out more work.
Recovery needs the image copy and the logs since it. DBRC is what ties them together; protect the RECON datasets.
Fragmented databases quietly slow every transaction. Watch the space and pointer statistics your DBAs report.
What you will see at work
- Operators and production support use /DISPLAY many times a day to answer 'is it running, and is anything queued?'
- DBAs own image copies, reorganisations and recovery jobs; developers are called in when bad data is involved.
- Disaster recovery tests prove that image copies, logs and RECONs can rebuild databases at the backup site.
Key terms
Check your understanding.
Take this lesson's quiz and save your progress. Free.