Long-running transactions

Do not design long-running transactions in such a way that they write frequently to the system log stream, in a single unit of work, across multiple activity keypoints.

If a long-running transaction does not take regular syncpoints while regularly initiating writes to the system log, CICS® is prevented from purging units of work that completed after the long-running task started.

In Figure 1, the long-running transaction is updating recoverable resources, causing writes to the system log. All the updates are in the same unit of work because it does not take explicit syncpoints. The oldest deletion point is earlier than activity keypoint number 5, preventing CICS from deleting any completed units of work that lie between activity keypoints 5 and 8. In this example, the long-running transaction will eventually cause the system log stream to spill onto secondary storage.
Figure 1. The effect of a ‘bad’ long-running transaction
This diagram shows the long-running transaction, described in the preceding text, spanning a number of activity keypoints numbered 5 through 9. At intervals, between the activity keypoints, the transaction performs updates to recoverable resources, but does not take a syncpoint. This forces CICS to retain all the log records and activity keypoints back to the start of the transaction, which is shown between keypoint 4 and 5.


dfht20u.html | Timestamp icon Last updated: Sunday, 4 June 2017