This topic describes the actions
that can occur when a node rejoins a cluster.
The following diagram describes the actions taken whenever a node rejoins
the cluster. In addition, the state of the rejoining nodes will be changed
from inactive to active in the membership
status field in the recovery domain of the CRG. The exit program is called
on all nodes in the CRG's recovery domain and is passed an action code of
Rejoin.
Table 1. Rejoin operationRejoin operation |
Rejoining node |
Cluster nodes |
Contains copy of CRG |
Does not contain copy of CRG |
Contain copy of CRG |
Do not contain copy of CRG |
(1) |
(2) |
(3) |
(4) |
Using the diagram above, the following situations are possible:
- 1 and 3
- 1 and 4
- 2 and 3
- 2 and 4
If a node in the cluster has a copy of the CRG, the general rule for rejoin
is that the CRG is copied from an active node in the cluster to the rejoining
node.
- Rejoin Situation 1
- A copy of the CRG object from a node in the cluster is sent to the joining
node. The result is:
- The CRG object is updated on the joining node with the data sent from
the cluster.
- The CRG object may be deleted from the joining node. This can occur if
the joining node was removed from the CRG's recovery domain while the joining
node was out of the cluster.
- Rejoin Situation 2
- A copy of the CRG object from the joining node is sent to all cluster
nodes. The result is:
- No change if none of the cluster nodes are in the CRG's recovery domain.
- The CRG object may be created on one or more of the cluster nodes. This
can occur in the following scenario:
- Nodes A, B, C, and D make up a cluster.
- All four nodes are in the recovery domain of the CRG.
- While node A is out of the cluster, the CRG is modified to remove B from
the recovery domain.
- Nodes C and D fail.
- The cluster is only node B which does not have a copy of the CRG.
- Node A rejoins the cluster.
- Node A has the CRG (although it is down level by now) and Node B does
not. The CRG is created on node B. When nodes C and D rejoin the cluster,
the copy of the CRG in the cluster updates node C and D and the previous change
to remove node B from the recovery domain is lost.
- Rejoin Situation 3
- A copy of the CRG object from a node in the cluster is sent to the joining
node. The result is:
- No change if the joining node is not in the CRG's recovery domain.
- The CRG object may be created on the joining node. This can occur if the
CRG was deleted on the joining node while cluster resource services is not
active on the node.
- Rejoin Situation 4
- Some internal information from one of the nodes in the cluster may be
used to update information about the joining node but nothing occurs that is
visible to you.