Logging

Logs are lists of activities, lists of work that have been done, and lists of changes that have been made. By knowing the state of the system when things were all right, and knowing what the system has done since then, you can recover it in the event of a failure.

Logs make recovery and restart possible. As IMS operates, it constantly writes its event information in log records. The log records contain information about:

• When IMS starts up and shuts down
• When programs start and terminate
• Changes made to the database
• Transaction requests received and responses sent
• Application program checkpoints
• System checkpoints

The following figure illustrates how logging works in an IMS online environment; logging in a batch environment is simpler.

Figure 1. Overview of the logging process

IMS externalizes log records by writing them to an Online Log Data Set (OLDS). To enhance performance and to optimize space on OLDSs, incomplete or partially filled buffers are written to a Write Ahead Data Set (WADS) when necessary for recoverability. The WADSs are high-speed DASD data sets with a high write rate. Only complete log buffers are written to OLDSs. When the log data is on an OLDS, the equivalent WADS records are ignored.

IMS uses a set of OLDSs in a cyclical way, which allows IMS to continue logging when an individual OLDS is filled. Also, if an I/O error occurs while writing to an OLDS, IMS can continue logging by isolating the defective OLDS and switching to another one. Once an OLDS has been used, it is available for archiving to a System Log Data Set (SLDS) on DASD or tape by the IMS Log Archive utility. The utility can be executed automatically through an IMS startup parameter (ARC=).

When IMS is close to filling the last available OLDS, it warns you so you can ensure that archiving completes for used OLDSs or add new OLDSs to the system. You can also manually archive the OLDSs to SLDSs using the IMS Log Archive utility. An SLDS can reside on DASD or tape. After an OLDS is archived, it can be reused for new log data. You use SLDSs as input to the database recovery process.

When you archive an OLDS, you can request that IMS write a subset of the log records from the OLDS to another log data set called the recovery log data set (RLDS). An RLDS contains only those log records required for database recovery.

While IMS is running and logging its activities, it takes periodic system checkpoints, and writes the checkpoint notification to another data set called the restart data set (RDS). IMS uses the RDS when it restarts to determine the correct checkpoint from which to restart.

In a batch environment, IMS writes log records directly to an SLDS, and does not use either the OLDSs or WADSs. Just as in the online environment, batch SLDSs can reside on DASD, tape, or mass storage. If the SLDS is on DASD, you can use an IMS utility to copy log records from DASD either to tape or to other DASD.