Dell SRDF Introduction

Preventing data loss

Should a failure occur, SRDF/Metro must ensure that there is no loss of data because of the failure. Also, SRDF/Metro decides which side that remains host-accessible (the winner). Having decided on the winner, the steps that SRDF/Metro takes to prevent data loss are:

1. Briefly stall all I/O operations to all devices on both sides of the SRDF/Metro group.
2. Hold I/O operations to both sides of all pairs in the SRDF/Metro group without processing them.
3. Set all devices on both sides of the SRDF/Metro group Not Ready on the SRDF link.
4. Make all devices on the loser side of the group inaccessible to the application host.

   Devices on with winner side remain accessible to the application host.

5. Clear the Stalled state.
   NOTE:The Stalled state is transient and typically transparent to applications. So, it does not appear as a discrete state, but is incorporated in the Suspended and Partitioned states.
6. Reject all the I/O operations that were held in step 2 with a retry sense code.

   The use of the retry sense code tells the host to send these I/O operations again.

Subsequent I/O operations are accepted by devices on the winner's side, and rejected by devices on the loser's side. This applies to all I/O operations, whether they are new or ones that are retried.

When the failure is resolved, SRDF/Metro:

1. Sets the devices in the SRDF/Metro group to Ready on the SRDF link.
2. Re-synchronizes each device pair in the SRDF/Metro group.
3. Sets each pair state to ActiveActive or ActiveBias once the pair can provide high data availability.