EMC VMAX 100K, Product Manual

Page: 139 / 200

Failback to the primary array
After the primary host and array containing the primary (R1) devices are again
operational, an SRDF failback allows production processing to resume on the primary
host.
Recovery for a large number of invalid tracks
If the R2 devices have handled production processing for a long period of time, there
may large numbers of invalid tracks owed to the R1 devices. SRDF control software
can resynchronize the R1 and R2 devices while the secondary host continues
production processing. Once there is a relatively small number of invalid tracks owed
to the R1 devices, the failback process can be initiated.
Temporary link loss
In SRDF/A configurations, if a temporary loss (10 seconds or less) of all SRDF/A links
occurs, the SRDF/A state remains active and data continues to accumulate in global
memory. This may result in an elongated cycle, but the secondary array dependent-
write consistency is not compromised and the primary and secondary array device
relationships are not suspended.
Transmit Idle on page 133 can keep SRDF/A in an active state during all links lost
conditions.
In SRDF/S configurations, if a temporary link loss occurs, writes are stalled (but not
accumulated) in hopes that the SRDF link comes back up, at which point writes
continue.
Reads are not affected.
Note
Switching to SRDF/S mode with the link limbo parameter configured for more than 10
seconds could result in an application, database, or host failure if SRDF is restarted in
synchronous mode.
Permanent link loss (SRDF/A)
If all SRDF links are lost for more than link limbo or Transmit Idle can manage:
l
All of the devices in the SRDF group are set to a Not Ready state.
l
All data in capture and transmit delta sets is changed from write pending for the R1
SRDF mirror to invalid for the R1 SRDF mirror and is therefore owed to the R2
device.
l
Any new write I/Os to the R1 device are also marked invalid for the R1 SRDF
mirror.
These tracks are owed to the secondary array once the links are restored.
When the links are restored, normal SRDF recovery procedures are followed
l
Metadata representing the data owed is compared and merged based on normal
host recovery procedures.
l
Data is resynchronized by sending the owed tracks as part of the SRDF/A cycles.
Data on non-consistency exempt devices on the secondary array is always dependent-
write consistent in SRDF/A active/consistent state, even when all SRDF links fail.
Starting a resynchronization process compromises the dependent-write consistency
until the resynchronization is fully complete and two cycle switches have occurred.
Remote replication solutions
SRDF recovery operations 139