Permanent suspend

If a node or a database IASP crashes with mainstore loss, the node will not be allowed to resume replication and will be marked as permanently suspended. If mainstore loss occurs, changes to replicated objects on that node may have occurred without being sent to the target node or tracked. Db2® Mirror cannot guarantee that all replicated objects are identical between the nodes.

The following figure shows that node B2PV has a replication state of BLOCKED. Its replication information indicates that it is permanently suspended.

Figure 1. Permanent suspend of node in Db2 Mirror GUI
Permanent suspend of node in Db2 Mirror GUI

This can also be seen using the QSYS2.MIRROR_INFO view.

Figure 2. Permanent suspend of node in MIRROR_INFO
Permanent suspend of node in MIRROR_INFO
The following situations can cause a node to become permanently suspended:
  • A partition crash with mainstore loss
  • Immediate power down of the partition from HMC
  • Transferring objects from an old load source disk to a new disk of a different size or type
The following situations can cause a database IASP to become permanently suspended:
  • Abnormal vary off of the IASP
  • A detach of the IASP while online if the IASP is the target of hardware replication, such as in a PowerHA® disaster recovery implementation
  • A hardware replication unplanned failover if the IASP is the target of hardware replication

Start of changeIf a node or IASP has a replication state of BLOCKED and the detailed information indicates that it is permanently suspended, the node or IASP must be recloned. It is recommended that the node which is permanently suspended should be the target (or setup copy node) of the reclone since some of its production data might have been lost when the node or IASP lost mainstore. If both nodes are permanently suspended, one node must be chosen as the source node. If the desired source node is not the primary node, the roles can be swapped prior to the reclone using either the Db2 Mirror GUI or the SWAP_MIRROR_ROLES procedure with the FORCE action specified. See Reclone process for information on performing a reclone.End of change

To avoid a permanent suspend:
  • Do not use the HMC to do an immediate power down of a Db2 Mirror node.
  • Take a mainstore dump (MSD) if a node crashes. The process of taking the MSD will write changed pages from mainstore to disk to avoid mainstore loss.
  • If an IASP is the target of hardware replication, such as in a PowerHA disaster recovery implementation, avoid a detach of the IASP while online.