The frequency with which consistency groups are formed is controlled by the number of seconds you specify for the Consistency Group Interval in the RSESSION START or RESUME command, plus the number of seconds it takes to actually form the consistency group. This sum, shown in Figure 1, is often referred to as the Recovery Point Objective (RPO). Point in time copies are written to the tertiary volume, or FlashCopy® target, according to the timing of the RPO.
The user-specified consistency group interval is the interval between the completion or failure to form a consistency group and the start of an attempt to form the next consistency group. It typically takes 3-5 seconds to form a consistency group, so if 0 is specified for the consistency group interval, the RPO would be 3-5 seconds, not 0. However, a consistency group interval of 0 causes consistency groups to be formed continuously, starting a new consistency group as soon as the previous one is formed. A peak in write I/Os can increase the amount of time it takes to form a consistency group because of increased drain time, the time it takes to update the remote secondaries with the coordinated primary updates. An insufficient bandwidth (due to a link failure or insufficient link configuration) can also increase the amount of time it takes to form a consistency group.
If a consistency group cannot be formed, data continues to flow to the secondary of the PPRC pair, but the point in time FlashCopy is not done.
A withdraw of a FlashCopy relationship within an Global Mirror session will impact the ability to form a consistency group. If you must withdraw a FlashCopy relationship, first remove the PPRC primary volume associated with the FlashCopy from the Global Mirror session with RVOLUME REMOVE, or pause the session with the RSESSION PAUSE command.
The suspension of a PPRC pair within a Global Mirror session will impact the ability to form a consistency group. If you must suspend a PPRC pair, first remove the PPRC primary volume from the Global Mirror session. If an error occurs causing a PPRC pair to be suspended, remove the pair from the session as soon as possible. Suspension of a PPRC pair within an active session will cause the next consistency group formation to fail.
As part of the consistency group formation, the master storage control must communicate with the subordinate storage controls to determine a consistent data point. A parameter called Maximum Coordination Interval (Figure 1) enables you to specify the maximum amount of time, in milliseconds, that the master storage control can spend communicating with storage control subordinates to determine a consistent data point. Since host write I/O is delayed while the consistent data point for a consistency group is being set, performance can be impacted when coordination cannot complete in an acceptable amount of time. If the Maximum Coordination Interval expires before formation of the consistent data point completes, the Consistency Group will be failed and the master will attempt to form the Consistency Group again after the next Consistency Group Interval time. The default value of the Maximum Coordination Interval is 50 milliseconds.
When a consistency group is formed, the data not yet copied must be transferred, or drained, to the PPRC secondaries at the remote site. During this time, new writes to the PPRC primaries are recorded in a change recording bitmap and not transferred to the PPRC secondaries. If one or more volumes in the consistency group have significantly more data to drain than the others, these volumes will prolong the time it takes to form the consistency group. This, in turn, can cause a backlog of updates on heavily updated primary volumes that must be transferred to the target volumes after the in-progress consistency group has been formed, resulting in an inefficient use of the PPRC links. A parameter called Maximum Consistency Group Drain Time (Figure 1) enables you to reduce the impact on the primaries by causing the consistency group to fail when the amount of time expires and the data to be transferred to the PPRC secondaries has not completed. At this point, all volumes can continue to transfer data to the remote site. The default value of the Maximum Consistency Group Drain Time is 30 seconds.
In order to make sure that a Consistency Group does get formed eventually, the Global Mirror master keeps track of how many times the Maximum Consistency Group Drain Time has consecutively expired. If it has expired 5 times (causing the Consistency Group formation to fail five times), the timers will be disabled and the Consistency Group will be allowed to form, no matter how long it takes unless there is a communications failure or fatal error impeding the drain.
A pause will not interrupt the formation of a consistency group. If the formation of a Consistency Group is in progress when a pause command is received, the pause will take effect after the formation of the consistency group is complete. The RSESSION STOP command can interrupt the formation of a Consistency Group in progress, unlike RSESSION PAUSE, which allows Consistency Group formation to complete. RSESSION STOP can interrupt a consistency group formation in process as long there is a consistent set at the recovery site. RSESSION STOP with the FORCE parameter will stop the session even if consistent data cannot be formed.