Planning disaster recovery actions

A common disaster recovery solution is to use Remote Copy technologies to replicate data on DASD volumes, where all the critical volumes from the primary site are mirrored to volumes at the disaster site. Among the critical volumes which can be mirrored in this way are those containing the Sysplex and CFRM couple data sets. Some special considerations apply to the volumes containing these couple data sets.

IBM® strongly recommends not to use synchronous remote copy technologies (such as peer-to-peer remote copy, or PPRC) for volumes that contain the Sysplex, CFRM couple data sets or both the Sysplex and CFRM couple data sets. Specifically, a DASD "freeze" against volumes containing these couple data sets could fatally interfere with the ability of the sysplex to recover and reconfigure itself during a site failure, connectivity failure, or whatever problem it was that triggered the DASD "freeze" to occur in the first place. The sysplex must be able to reconfigure itself when necessary, and during synchronous remote copy failover processing, this is not possible.

Start of change IBM also recommends against the use of asynchronous remote copying technologies such as Global Mirror for managing couple data sets. Asynchronous copy transfers storage unit track images that do not necessarily correspond to the logical records defined within the couple data set. After failover to the remote site with which the copy was established, it may be impossible to IPL the sysplex using the copied couple data sets. End of change

Start of change Instead, IBM recommends that you predefine and initialize new instances of couple data sets to be used for sysplex initialization at the disaster recovery site. In this case, a different COUPLExx parmlib member should also be defined for use at the disaster recovery site, which contains the couple data set names to be used at the disaster recovery site, and the name of the CFRM policy that is to be activated for use at the disaster recovery site. End of change

Note:

When initializing the sysplex with freshly-formatted couple data sets, the CFRMPOL keyword in the COUPLExx parmlib member can be used to identify the desired CFRM policy that is to be activated automatically when the sysplex is initialized at the disaster recovery site.
In general, the CFRM policy that is used at the disaster recovery site is not the same policy that is normally used for production. For example, a different set of coupling facility images, with unique physical identification information, exist for use at the disaster recovery site rather than exist in the production site; the CFRM policy used at the disaster recovery site needs to define those coupling facility images for use, not the production CFs. The set of required structure definitions might also be different when running at the disaster recovery site. You can either define two totally separate CFRM policy definitions for this (one for use at the production site, and one for use at the disaster recovery site), or define one "overdefined" CFRM policy for use at both sites, which encompasses all the CF definitions and CF structure definitions that are applicable to both sites.
When IPLing with freshly-formatted sysplex couple data sets, operator intervention may be necessary at several points during initialization of the first system. Some applications retain configuration information in the user data portion of the records associated with their groups in the sysplex couple data set. This information will not be available when IPLing with new couple data sets, and may have to be reestablished by interaction with the operator.
GDPS customers should refer to GDPS procedures for couple data set management, particularly for the Logger couple data set.

Important!

Great care should be taken when replicated Sysplex or CFRM couple data sets are used to initialize a sysplex for disaster recovery testing, whenever those couple data sets are replicated by any of the following means:

Flash copy
Synchronous remote copy
Asynchronous remote copy
Having the primary couple dataset located in one site and the secondary couple data set in the other site
Using an "old" couple dataset that was formerly in use in the production sysplex

The replicated couple data sets will contain all of the internal sysplex information from the production sysplex, which includes the following:

The identical sysplex name
Active system names
Coupling facility names
Resource ownership information

When a sysplex is initialized at the disaster recovery site using these replicated couple data sets, that sysplex will work as if it is the production sysplex. If there is physical connectivity between the disaster recovery site and the production site that allows this new instance of the sysplex to access the physical resources of the production sysplex which are still being used by the production sysplex, serious problems can result as both sysplexes erroneously attempt to share and manage the same physical resources, such as coupling facilities and CF structures.

Such problems cannot occur when the sysplex that is initialized at the disaster recovery site uses freshly-formatted Sysplex and CFRM couple data sets, and uses the CFRMPOL keyword in the COUPLExx parmlib member. The CFRMPOL keyword in the COUPLExx parmlib member can automatically initialize a disaster-recovery-only version of the CFRM policy that only defines and references resources that are valid in the disaster recovery site instead of the production site.

Figure 1 shows an example of using the CFRMPOL keyword in the COUPLExx parmlib member.

Figure 1. Example: specifying the CFRMPOL keyword in COUPLExx parmlib member

COUPLE    SYSPLEX(&SYSPLEX.)
          CFRMPOL(CFPOLD)
          PCOUPLE(SYS1.XCFCDS.PRI)
          ACOUPLE(SYS1.XCFCDS.ALT)
          INTERVAL(25)
          CLEANUP(10)
          MAXMSG(2500)

CLASSDEF  CLASS(SMALL) CLASSLEN(956) GROUP(UNDESIG)
CLASSDEF  CLASS(DEFAULT) CLASSLEN(4028) GROUP(UNDESIG)
CLASSDEF  (LARGE) CLASSLEN(16316) GROUP(UNDESIG)

PATHOUT   STRNAME(IXCSTR1) CLASS(DEFAULT)
PATHOUT   STRNAME(IXCSTR2) CLASS(SMALL)
PATHOUT   STRNAME(IXCSTR3) CLASS(LARGE)
PATHOUT   STRNAME(IXCSTR4) CLASS(SMALL)
PATHIN    STRNAME(ISCSTR1,IXCSTR2,IXCSTR3,IXCSTR4)

DATA      TYPE(CFRM)
          PCOUPLE(SYS1.CFRMCDS.PRI)
          ACOUPLE(SYS1.CFRMCDS.ALT)