Plan DRXRC-type staging data sets for coupling facility log streams

The use of DRXRC-type staging data sets provides DASD mirroring capability for Logger exploiters that previously could not take advantage of duplexing to log stream staging data sets. System logger and Extended Remote Copy/System Data Mover (XRC/SDM) provide support for remote mirroring of high volume logging applications, such as IMS™ CQS and CICS®. This support allows you to choose asynchronous writes to staging data sets for log streams. Previously, all writes had to be synchronous with respect to writing data to a Coupling Facility structure before responding to the log stream writer and this limited the throughput on the primary systems for high-volume logging applications. The ability to do asynchronous writes to the log stream staging data sets can allow the use of XRC as a mirroring technology for some applications for which it was not previously practical.

DRXRC-type staging data sets are used when a log stream is defined with STG_DUPLEX(YES),DUPLEXMODE(DRXRC) and will be used for additional log data duplexing in order to provide remote disaster recovery opportunity. See the DUPLEXMODE keyword description in Selecting a method of duplexing coupling facility log data and LOGR keywords and parameters for the administrative data utility for additional details on the use of this specification.

Using DRXRC-type staging data set use provides DASD mirroring capability for Logger exploiters that previously could not take advantage of duplexing to log stream staging data sets because of log data write throughput performance considerations; meaning it was too costly for the I/O overhead per IXGWRITE to the log stream. By allowing the MVS™ system logger staging data sets to be written to asynchronously and providing the appropriate consistency group times on a recovery site, Logger and XRC provide a viable mechanism for an enterprise to use a long distance disk mirroring solution to recover log stream data written to a coupling facility structure in the event of a disaster at their primary sysplex site.

Basically, system logger exploiters continue to have a mechanism to use log stream staging data sets and DASD mirroring technology to provide remote site disaster recovery capability for coupling facility based log streams. The existing XRC capabilities, combined with the asynchronous writing to log stream data sets, provides a more complete method of mirroring log data written into a coupling facility structure. Peer to Peer Remote Copy (PPRC) or any non-XRC (LOGPLUS) configurations will not provide the correct environment for the proper use of DRXRC-type staging data sets.

Note: An installation can continue to duplex log data to log stream staging data sets for recovery purposes at their primary sysplex site in conjunction with the new DRXRC-type staging data sets, but only one type of duplexing specification is allowed for a given log stream. The log streams using conventional staging data sets will still be used for log data recovery at the primary site or at a secondary recovery site. Also, note when the DRXRC option is used to duplex log data, careful consideration needs to be given to the primary (local) systems duplexing and log data recovery capabilities to ensure both the throughput and recoverability aspects are met for the log stream.

In addition, IBM® recommends you NOT define a log stream to use the DUPLEXMODE (DRXRC) attribute in the same CF structure as other log streams that will be connected on pre-z/OS V1R7 release level systems. Logger's log stream log data recovery processing during system level recovery or following structure rebuilds can result in any system connected to the structure attempting to recover for any log stream in the structure. If log stream recovery is attempted on a pre-z/OS V1R7 release level system for log streams with the DRXRC attribute, the following undesirable effects may result:

IXG212E OR IXG213E messages indicating the log stream recovery was not successful.
The log stream in question will most likely be marked with a possible loss of data condition.
The DRXRC-type staging data set may be deleted and not available for appropriate use on z/OS® V1R7 level systems.

There are specific DASD and configuration requirements when using XRC and Logger DRXRC-type staging data sets. See z/OS DFSMS Introduction, z/OS DFSMS Advanced Copy Services, DFSMS Advanced Copy Services Installation Planning Guide, GC35–0481, and DFSMS Extended Remote Copy Reference Information for Advanced Users, GC35–0482, for more information about XRC.

IBM recommends using IBM TotalStorage Enterprise Storage Server® (ESS) when using DRXRC-type staging data sets since it is a high capacity storage subsystem that can be configured according to your requirements. The ESS function authorization needs to include XRC capability as an available feature.
The installation storage administrator needs to establish the appropriate SMS class, group definitions and/or automatic class selection (ACS) routines to manage, or at least assign the constructs, the log stream DRXRC-type staging data sets.
Configure the volumes in a manner that allows only one DRXRC-type staging data set per volume. The installation must ensure NOT to allow any other data sets to be placed on a volume with a DRXRC-type staging data set.
IBM recommends sizing the volume according to the DRXRC-type staging data set size (see below).
Define the appropriate log streams to make use of the DRXRC-type staging data sets by specifying the STG_DUPLEX(YES),DUPLEXMODE(DRXRC) attributes.
The log stream STG_xxxxCLAS and STG_SIZE attributes need to be specified appropriately to ensure each DRXRC-type staging data set is sized correctly and mapped to its own volume.
Since a single volume is used for each DRXRC-type staging data set, VSAM data striping cannot be used for these data sets.

Also note, Logger only supports any type of staging data set up to 2 gigabytes in size.

If IBM TotalStorage Enterprise Storage Server is used for DRXRC staging data sets, then IBM recommends sizing the DRXRC-type staging data set as large as the size of maximum coupling facility structure space allocated for the log stream. This recommendation is possible using ESS as it allows configuration of proper size volume.

For fixed size storage devices (for example, mod-3 3390), IBM recommends sizing 2GB for DRXRC staging data sets. This means using STG_SIZE (524288) on the log stream definition to avoid having more than one staging data set on a single volume.
All the volumes destined to contain DRXRC-type staging data sets need to be defined as part of an Extended Remote Copy (XRC) storage control session pair. On the XADDPAIR command or ANTRQST XADD API, specify the LOGPLUS keyword to indicate that the primary volume of the volume pair is to be explicitly written to by Logger when it is duplexing to its DRXRC-type staging data set. See the XADD parameters (e.g. DONOTBLOCK or DVCBLOCK) pertaining to minimizing the impact to write response times for these high activity (LOGPLUS) sessions.
For each storage control session (scsession), as defined through XADDPAIR command, a utility device is also required to read updates to the primary volumes belonging to that particular scsession. XADDPAIR command is also used to define the utility device for each LOGPLUS pair.
```
XADDPAIR TEST VOLUME (XRCP00 XRCS00 XRCP2E XRCUTL) QUICKCOPY LOGPLUS
```
The actual physical session ids of the LOGPLUS scsessions will start at X'FF' and go downwards.
When a Volume Pair has been added by using the LOGPLUS keyword, XQUERY reports will show scsessions as:
L+

when system logger is actively using the volume, or

L-

when system logger is not actively using the volume.
Consider which XRC ERRORLEVEL specification on the XADDPAIR or XSTART commands to use involving LOGPLUS volumes. ERRORLEVEL keyword indicates the level of detection and action to use if the XRC (logical) session consistency group time is compromised.
Only ERRORLEVEL(SESSION) preserves consistency across all volumes in the session. Use of VOLUME or group_name ERRORLEVEL will allow session consistency time to advance for error conditions even for L+ volumes.
Add all the necessary volumes that contain the Logger infrastructure consisting of at least the log stream offload data sets, staging data sets other than DRXRC type, LOGR couple data sets, catalogs, etc., and volumes used by Logger exploiters to the same XRC (logical) session (using XADDPAIR commands) mentioned in the prior step.
Note: Do not specify the LOGPLUS keyword on the XADDPAIR commands for these volumes.
If you make use of multiple XRC storage control sessions with the LOGPLUS designation or some with and some without this designation, then couple the different sessions into a single master XRC session. See the XCOUPLE command for details. If multiple XRC sessions are being used in a Logger configuration with LOGPLUS designated volumes and the XRC sessions are not coupled into a master session, then system logger will not be able to correctly manage log data on the recovery site.
All LOGPLUS volumes must be included within the same XRC session or at most one master Coupled XRC session. Other master Coupled XRC sessions can be used, but they will not be able to contain additional LOGPLUS volumes in order for the Logger DRXRC-type staging data set recovery to work.
The above configuration requirements will result in using one Logger DRXRC-type staging data set per volume per XRC storage control session (scsession). This 1:1:1 ratio could impact the overall XRC and system resource configuration, especially if you choose to mirror large number of log stream staging data sets. In that case, see z/OS DFSMS Advanced Copy Services for more planning XRC/SDM when using LOGPLUS volumes.
In order for Logger to acquire the consistency time for the DRXRC-type staging data set in the recovery site, add new XRC parmlib category CONTIME, and new parameters DefaultSessionId and DefaultHlq to the XRC parmlib member. If the LOGPLUS volumes are included in an XRC master session, then use these parameters to identify the XRC master session. If the LOGPLUS volumes are wholly contained within one XRC session, then use these parameters to identify that particular XRC session.
These XRC parmlib values may be dynamically changed by the customer using the XSET command. Normally the parmlib values for the control address space (ANTAS000) may not be changed with the XSET command since the control address space does not have a session id. When a session id of ANTAS000 is used, then XSET will change the parmlib values for the control address space. If the CONTIME member is not defined, users still have the option to use the current time as the consistency time for log data.
Note, the IBM Device Support Facilities ICKDSF volume utility commands INIT and REFORMAT (XRCLOGGER parameter) aids in managing LOGPLUS designated volumes. See Device Support Facilities (ICKDSF) User's Guide and Reference. The ICKDSF utility may be needed if a volume state is left as L+ even after Logger is no longer actively using the volume.

Any log stream DRXRC-type staging data set will only be usable for recovery purposes when a system is IPLed with the DRMODE=YES system parameter specification. This type of IPL would normally be done when the first secondary site recovery system is initially IPLed to handle the start up following a disaster situation for the primary (main) sysplex systems. It is assumed that the necessary actions (e.g. XRC XEND and XRECOVER commands) had been taken to establish the DASD consistency groups related to the Logger data sets before a recovery system being IPL'd with DRMODE=YES. See System logger recovery and DRXRC considerations and DRMODE=YES IPL option for more details on how to make DRXRC-type staging data sets available for log data recovery.

For DRMODE=YES specific IPLs, Logger will issue message IXG068D indicating that a decision be made by the installation to confirm (Y) or ignore (N) the DRMODE=YES IPL system parameter specification. Before responding to message IXG068D, enter a D LOGGER,L,LSN=* command and view the resulting IXG601I message to identify any log streams that have "STAGING DRXRC" included as a duplexing method. These log streams are the ones that have DRXRC-type staging data sets and will be affected by the response to the IXG068D message.

A reply of Y to message IXG068D will cause Logger to issue message IXG072I indicating the consistency time that will be used by Logger for log data associated with the DRXRC-type staging data sets. Logger will also make use of log stream DRXRC-type staging data sets eligible for log data recovery purposes.

To prepare log streams for recovery in this situation, the log streams that had DRXRC-type staging data sets in use are identified, and these log streams will have their attributes updated to STG_DUPLEX(NO),DUPLEXMODE(). Message IXG224I is issued, to hardcopy, for each log stream that is updated. This attribute change is made since the intent of DRXRC-type staging data sets is to allow the log stream recovery to occur on the secondary (remote) site. Once the log data has been recovered, the installation's need for using DRXRC-type staging data sets had just been satisfied. It is expected that continued use of these recovered log streams on the secondary sysplex (recovery) site should not automatically cause new staging data set duplexing. The installation will have to reestablish any specific duplexing if the automatic result of log stream duplexing is not desired.

After the appropriate log streams have their DRXRC-type staging data sets marked as eligible to be included for log data recovery, message IXG069I is issued and Logger will perform its system level recovery for log streams that have failed connections for the system being IPLed.

Once message IXG069I is issued, all remaining recovery systems can be IPLed with DRMODE=NO. system logger will use the previously defined DRXRC-type staging data sets as part of the system level log stream recovery on those systems as well. If message IXG069I has not been issued a new message, IXG073D will be issued indicating that Logger has failed to mark the DRXRC-type staging data sets for log data recovery. Installation will be asked to reply 'R' to retry and 'C' to cancel the conversion phase. Before replying to the IXG073D message, enter a D LOGGER,L,LSN=* command and view the resulting IXG601I message to identify any log streams that have STAGING DRXRC included as a duplexing method. If 'C' is replied as a response to message IXG073D then data for all the log streams, with duplex mode DRXRC, will not be recovered.