HP 3PAR Remote Copy Software - Manual (Page 55)

2.39 Error Handling
InForm OS Version 2.3.1 Remote Copy User’s Guide
■ If the no_fail_wrt_on_err policy is enabled for a synchronous volume group (this is the
default), the host I/O completes without error. Replication is stopped for the volume group
and a snapshot is taken of each volume in the volume group. In addition, the offset and
length of the I/O that failed is recorded in nonvolatile memory. An alert is generated to
indicate the I/O error on the secondary storage server and to report the stopping of the
volume group.
The failed I/O needs to be recorded separately because the snapshot is taken after the I/O
was written on the primary volume but before it could be completed on the backup
volume. When replication is restarted for the volume, the failed I/O is read from the
primary volume and sent to the backup. In addition, all differences between the base
volume and the snapshot taken when the replication was stopped are also sent over to
resynchronize the backup volume with the primary volume.
■ If the fail_wrt_on_err policy is enabled for a synchronous volume group, then a write
error is returned to the host when either the primary or the backup volume reports an I/O
error, but the replication is not stopped. If the host retries the failed write and subsequent
host writes continue to encounter errors while writing to the backup volume, errors
continue to be returned to the host.
■ auto_recover does not automatically recover from stopped groups as a result of write
errors. It only automatically recovers from link failures.
2.13.6 Read Errors
During the initial synchronization of a backup volume with its primary volume, Remote Copy
either reads the base volume (for synchronous volume groups) or creates a read-only snapshot
of the base volume (for asynchronous periodic volume groups). If a read request fails for any
reason, the synchronization process cancels and the secondary volume is marked NotSynced .
An alert is issued, indicating that the synchronization has failed. However, this does not stop
the replication. To get the volumes back in synchronization, manually issue the syncrcopy
command for synchronous volume groups as well as asynchronous periodic groups.
When a synchronous volume group is stopped and started after it has been fully synced, a
resynchronization is automatically started. This process compares a read-only snapshot taken
when the group was stopped with the current base volume. Any changed pages are read from
the base volume and sent over to the secondary. This type of read failure also results in the
cancellation of the resynchronization process with the secondary volume marked NotSynced .
The group is not stopped. To recover from this, you can manually issue the syncrcopy
command.