HP XP P9500 Storage - Manual (Page 166)

Notice on system operation
RAID Manager does not change the command options for supporting Continuous Access Journal
MxN Open. However, the output of the command is added so that the command can display the
CT group and Q-Marker for each storage system, because the CT group and Q-Marker are managed
on each storage system.
(1) CT group ID and journal ID for Cnt Ac-J
The CT group ID must be assigned/used as the unique identifier across multiple storage systems.
Therefore, the paircreate command makes a group volume having the same CT group ID across
multiple storage systems.
(2) Cascading operation
The "-FHORC" option for cascading operation does not perform with CTQ-Marker Mode. Hence
the cascading volume must not be specified Cnt Ac-J MxN volume when using the "-FHORC"
option.
(3) Running HORCM daemon
HORCM daemon process delivers the timestamps (called the consistency Q-Marker) to the defined
CT group. Therefore if HORCM daemon is stopped, then Cnt Ac-J will stop to delta (commit) and
will suspend because of Journal Full state.
Thus HORCM daemon must be running to keep the PAIR state.
(4) Separating a command device IO from application IO
The HORCM daemon process issues IOs to the command device in order to deliver the timestamps
to the defined CT group. Therefore, it is recommended to separate the command device path from
the application IO path.
(5) About waiting application IO
The HORCM daemon process delivers the timestamps to the defined CT group while freezing IO
for each journal.
Waiting Rate = 0.5ms * Number of journal / / Interval (ms) * 100
(Note that 0.5 ms depends on the OS platform.)
Therefore it is recommended to limit within four journal per CT group and 8192 LDEVs per CT
group. In the case of multiple CT groups per HORCM, it is recommended to limit within 256 LDEVs
per CT group.
(6) HOST IO on pairsplit -RS, -R, horctakeover, pairresync -swapp(s)
The pairsplit -RS, -R operation cannot be frozen and Split HOST IO from RCU in Link
Normal state. In other words this option does not support At-time Split, hence these operations are
required to stop HOST IO in order to keep Data Consistency on S-VOL. This is the same restriction
as Cnt Ac-S_Sync.
(7) Suspending/deleting status
The suspending/deleting for committing S-VOL Journal data with CTQ-Marker is accomplished by
the RAID Manager commands. Therefore the storage system has nothing to do in that status
(suspending/deleting).
If a RAID Manager command is aborted for some reason (KILL, etc.) or EX_EWSTOT, the storage
system keeps that status (suspending/deleting).
To terminate this status, re-execute the RAID Manager command, then terminate suspending/deleting
status after "JNL Consistency Restore".
166 Data replication operations with RAID Manager